PP 2 Man

Pāli
PLATFORM 2
The Official Manual
J. R. Bhaddacak
Version 2.0
Copyright © 2023 J. R. Bhaddacak
This work is licensed under a Creative Commons

Attribution-NonCommercial-NoDerivatives 4.0 International License.
https://creativecommons.org/licenses/by-nc-nd/4.0/
Release History
Version Built on Description
2.0 10 Nov 2022 First release, bundled with the program
Preface
Pāli Platform 1 has been released since January 2020. It might

be only me who used it substantially to produce Pāli learning
books.1 After I finished the books, I have spent many months
(9 or so) to rewrite Pāli Platform, to make it modernized in
look, more user-friendly, to add more features, and to fix many
bugs.2 Now its version 2 has been done. And I think it is the
time to write a full-blown user’s manual because the program
is quite sophisticated now. So here we are.
After the release of Pāli Platform 2, the old version will be
discarded. So, the users have to learn a new way for doing
things, for example, methods of typing in Pāli characters. We
will learn about all features in due course. However, I still
cannot explain every bit of details. One reason for this is the
graphic user interface of the program is intuitively easy to
1 Pāli for New Learners, Book I and II
2I moved from Java 8 to Java 11 and replaced Swing UI with JavaFX. I
thought at first I would use Kotlin, but it is still more difficult to work with.
Java is still the best friend I have. It is not the language itself is great, but the
whole Java ecosystem makes programming with Java enjoyable: enormous
available libraries, fast compile time, excellent API documentation, flexible
and uncomplicated deployment scheme, etc. Moreover, as I work with JavaFX
so far, this GUI is the best among all cross-platform GUI toolkits. It can run
and look the same in all platforms without any adjustment. Some might say
Java is slow, at least in comparison with C or C++. As you shall see by using
the program, it just needs start-up time. After that the slowness is negligible.
iii
learn, and I have already provided some helps in several places.
Still, the users need to spend time playing with the program
to gain familiarity.
I divide the manual into five parts. The first part, Essen-
tial Starter, is crucial. It is supposed to be read deliberately.
The main purpose of this part is to help the users start the
program successfully in various contexts, and to provide an
initial guidance and troubleshooting.
The second part is all about grammatical tools. It is enough
to just go through the part quickly and come back when neces-
sary. Several chapters are short. The longest one is Chapter
9 (Prosody), which needs an elaborate treatment.
The third part is about the Pāli collection. You will learn
how to find a document and open it, how to use the viewer,
and how to deal with the overall term list. This part is more
substantial than the previous one and each chapter is not long.
So, it should be read carefully.
The fourth part is about advanced search tools. It has one
big chapter about Lucene Finder. This chapter needs a care-
ful read because of the complexity of its functions. Another
chapter, about Tokenizer, is short, not because it is simple, but
rather it requires the understanding of the previous chapter.
The tool itself is quite complicated and needs an exploration
by the users themselves.
The fifth part is the rest of the all above. You will learn var-
ious tools, like the program’s text editor, batch script trans-
former, and more importantly the text Reader. The Reader,
together with its companion tool, Sentence Manager, is an in-
novative tool that can help the learners read Pāli texts more
conveniently. Furthermore, translations can be added to the
texts at sentence level by this set of tools.
The last chapter is a short treatment of regular expression.
Because I add this search function to the program in various
places, some guidance is needed. For the topic itself is big
and beyond our main concern, what I can do is just a survival
introduction.
My target readers of this manual are those who want to
make use of Pāli Platform to its full capacity, both for learn-
ing and researching purpose. Some basic knowledge of Pāli is
helpful, particularly the terminology used in the field. For the
fundamental of the language, see the books mentioned. Users
iv
Preface
outside the sphere of computational technology may skip com-

puter-related technical terms that appear unexplained, or else
they can surely find an explanation in Wikipedia.
This manual is a little hastily written. It is a product of
one and a half month of my full-time working.3 My intention
is to bundle the manual with the program, and release all of
them before the end of 2022. The program has been tested and
debugged along the way when I have written the book. So, ev-
erything should look nearly complete. Yet, errors always lurk
somewhere to show up, both in the books and in the program.
If you find something, or many things, unusual, please kindly
report it to me.4
3 I usually work 5–6 hours a day, no weekend. I use only one day in a week
to connect to the Internet, mostly for updating information and searching for
needed materials. Yet, I still work 2–3 hours that day. I choose to make this
manual with LATEX, despite its laborious process of writing, because it looks
authoritative and it is easier and more pleasurable to read.
4 jakratep at gmail dot com
v
Contents
Preface iii
Contents vii
List of Tables x
List of Figures xi
I. Essential Starter 1
1. Begin at the beginning 3

1.1. A history of Pāli Platform . . . . . . . . . . . . . 3
1.2. Why I take Pāli seriously? . . . . . . . . . . . . . 4
1.3. Features so far . . . . . . . . . . . . . . . . . . . . 6
1.4. How to run the program . . . . . . . . . . . . . . 7
1.4.1. Windows . . . . . . . . . . . . . . . . . . . 9
1.4.2. GNU/Linux . . . . . . . . . . . . . . . . . . 10
1.4.3. macOS . . . . . . . . . . . . . . . . . . . . 11
1.5. When things go right . . . . . . . . . . . . . . . . 12
1.6. When things go wrong . . . . . . . . . . . . . . . 13
1.7. Download links . . . . . . . . . . . . . . . . . . . 14
vii
Contents
2. Basic operations and settings 16

2.1. Main window . . . . . . . . . . . . . . . . . . . . . 16
2.2. Common tool bar . . . . . . . . . . . . . . . . . . 17
2.3. Fonts and problems . . . . . . . . . . . . . . . . . 18
2.4. Pāli input . . . . . . . . . . . . . . . . . . . . . . . 19
2.5. Minor concerns . . . . . . . . . . . . . . . . . . . 21
II. Grammatical Tools 23
3. Dictionaries 24
4. Letters 29
5. Declension table 31
5.1. Pronouns . . . . . . . . . . . . . . . . . . . . . . . 31
5.2. Nouns/Adjectives . . . . . . . . . . . . . . . . . . 33
5.3. Numbers . . . . . . . . . . . . . . . . . . . . . . . 34
6. Verbs 36
7. Conjugation table 38
8. Roots 41
9. Prosody 43
9.1. A survival introduction to Pāli prosody . . . . . . 44
9.2. Two types of prosodic patterns . . . . . . . . . . 45
9.3. Verse types of mattāvutti . . . . . . . . . . . . . . 48
9.4. Verse types of vaññavutti . . . . . . . . . . . . . . 53
III. Pāli Collection 59
10. Browsing and bookmarking 60
11. Document Finder 64
12. Document viewer 67
13. Simple Lister 72
viii
Contents
IV. Advanced Search Tools 78
14. Lucene Finder 79

14.1. Options for indexing . . . . . . . . . . . . . . . . . 80
14.2.Description of fields . . . . . . . . . . . . . . . . . 82
14.3.Lucene simple search . . . . . . . . . . . . . . . . 83
14.4.Lucene query syntax . . . . . . . . . . . . . . . . 85
14.4.1. Using wildcards . . . . . . . . . . . . . . . 86
14.4.2. Using regular expression . . . . . . . . . . 86
14.4.3. Using fuzzy query . . . . . . . . . . . . . . 86
14.4.4. Using proximity . . . . . . . . . . . . . . . 87
14.4.5. Using range . . . . . . . . . . . . . . . . . 87
14.4.6. Using term boost . . . . . . . . . . . . . . 88
14.4.7. Using logical operators . . . . . . . . . . . 89
14.5.Concluding remarks . . . . . . . . . . . . . . . . . 90
15. Tokenizer 91
V. Miscellaneous Tools 94
16. Pāli Text Editor 95
17. Batch Script Transformer 97
18. Pāli Text Reader 99
19. Sentence Manager 105
20. Quick guide to regular expression 109
About the author 113
Colophon 114
ix
List of Tables
5.1. Expansion of case abbreviations . . . . . . . . . . 32
7.1. Tenses and moods . . . . . . . . . . . . . . . . . . 40
9.1. Syllable vs. weight summation . . . . . . . . . . 46

9.2. Meter groups used in mattāvutti . . . . . . . . . 46
9.3. Meter groups used in vaññavutti . . . . . . . . . 47
9.4. Syllable vs. weight grouping . . . . . . . . . . . . 48
9.5. Symbols in verse formulas . . . . . . . . . . . . . 48
9.6. Verse types of mattāvutti . . . . . . . . . . . . . . 49
9.7. Verse types of vaññavutti . . . . . . . . . . . . . . 54
12.1. Transformation rules of Thai script . . . . . . . . 71
14.1. Fields used in Lucene Finder . . . . . . . . . . . 82
20.1. Some uses of regular expression . . . . . . . . . . 110
x
List of Figures
1.1. PaliPlatform2’s folder structure . . . . . . . . . . 8

1.2. Main window of Pāli Platform 2 . . . . . . . . . 12
1.3. Information of runtime environment . . . . . . . 13
1.4. An example of the JVM’s error message . . . . . 14
2.1. The common local tool bar . . . . . . . . . . . . . 17

2.2. Settings for Pāli input . . . . . . . . . . . . . . . . 20
3.1. Menu Grammar . . . . . . . . . . . . . . . . . . . . 24

3.2. Dictionaries window . . . . . . . . . . . . . . . . 25
3.3. Default dictionary selection . . . . . . . . . . . . 26
3.4. Dictionaries’ wildcard search . . . . . . . . . . . 27
3.5. Dictionaries’ meaning search . . . . . . . . . . . 28
3.6. Dictionaries’ result opened in an editor . . . . . 28
4.1. Letters window . . . . . . . . . . . . . . . . . . . 30
5.1. Declension table of a pronoun . . . . . . . . . . . 32

5.2. Declension of tumha against term list . . . . . . 33
5.3. Declension of vimutti against term list . . . . . . 33
5.4. Declension of sara . . . . . . . . . . . . . . . . . . 35
5.5. Declension of 100250 . . . . . . . . . . . . . . . . 35
xi
List of Figures
6.1. Result of main verbs when search ‘vim’ and se-

lect vimuccati . . . . . . . . . . . . . . . . . . . . 36
6.2. Options of other verb forms . . . . . . . . . . . . 37
7.1. Aorist forms of pacati . . . . . . . . . . . . . . . . 39

7.2. Masculine ta forms of pacati . . . . . . . . . . . . 40
8.1. Roots window . . . . . . . . . . . . . . . . . . . . 42
9.1. Prosody window . . . . . . . . . . . . . . . . . . . 43

9.2. Meter calculation in the program’s editor . . . . 45
9.3. Analysis of Ariyā verse type . . . . . . . . . . . . 51
9.4. Example of an over-required case . . . . . . . . 52
9.5. Analysis of Tanumajjhā verse type in edit mode 57
10.1. TOC Tree window . . . . . . . . . . . . . . . . . . 61

10.2.TOC Tree window at text level . . . . . . . . . . 62
10.3.Bookmarks window . . . . . . . . . . . . . . . . . 63
11.1. Searching Dhammapada in Document Finder . . 65

11.2.Searching with a wildcard in Document Finder . 65
11.3.Content searching in Document Finder . . . . . 66
12.1. Document viewer in its full form . . . . . . . . . 68

12.2.Document viewer with Quick Dictionary . . . . . 69
12.3.Document viewer displaying in Myanmar script 70
12.4.General Settings of script transformation . . . . 70
13.1. Simple Lister window with term summary . . . 73

13.2.Top longest terms in Simple Lister . . . . . . . . 75
13.3.Filter by meter ‘gggl’ in Simple Lister . . . . . . 76
13.4.Filter by meter ‘2221’ in Simple Lister . . . . . . 76
13.5.Grouping by the first letter in Simple Lister . . . 77
13.6.Grouping by the last 2 letters in Simple Lister . 77
14.1. Apache Lucene’s logo . . . . . . . . . . . . . . . . 79

14.2.Options for building Lucene index . . . . . . . . 80
14.3.Simple one-term search in Lucene Finder . . . . 83
14.4.Simple two-term search in Lucene Finder . . . . 84
14.5.Proximity search in Lucene Finder . . . . . . . . 87
14.6.Range search in Lucene Finder . . . . . . . . . . 88
14.7. Term boosting in Lucene Finder . . . . . . . . . . 89
xii
List of Figures
15.1. Tokenizer window in full . . . . . . . . . . . . . . 92
16.1. Text-processing tools in Pāli Text Editor . . . . . 96
17.1. Batch Script Transformer window . . . . . . . . 97
18.1. A sentence with translations in Pāli Text Reader 101

18.2.Detail mode in Pāli Text Reader . . . . . . . . . . 103
18.3.A use of edit in Pāli Text Reader . . . . . . . . . 103
19.1. Sentences tab in Sentence Manager . . . . . . . 106

19.2.Translation Variants tab in Sentence Manager . 107
19.3.Merger tab in Sentence Manager . . . . . . . . . 108
xiii
Part I.
Essential Starter
1
Begin at the beginning
1
In this starting chapter, I will tell you a short story of the de-
velopment of the program. At the end, we will learn how to
make it run in your computer.
1.1. A history of Pāli Platform

Originally, I had an idea to make a program that can search
words in the Pāli canon effectively, at least more effective than
the programs we had at the time. That was around 2–3 decades
ago, when I was interested in Buddhism after I finished my
undergrad engineering program. At the time, it was too dif-
ficult to do with limited technology and data available, and I
had many more pressing things to do. So, the idea just got
lingered in my mind since.
After I found that we have a digital version of the whole
Pāli collection, distributed by Vipassana Research Institute
(VRI) via tipitaka.org, the possibility of the project began to
light up. But I still had more important things to do. Once
I quit my job, I had an opportunity to continue my studies.
Then I finished a master program in Software Engineering.
The subject of Information and Retrieval (IR) was the main
concern of that study.1
1 Meanwhile, I also studied Buddhism and Philosophy at another univer-
3
1. Begin at the beginning
Nothing still happened yet. My life changed. In 2010, I be-

came a monk and have been enjoyed the peaceful life. The
idea has been neglected. At some point, I thought it is bet-
ter for me to deepen my studies. Then in 2018, I finished a
PhD program in Religious Studies. After that, I had time and
nothing important to pursue. So, I started the Pāli Platform
project, and finished its first version in January 2020, by one
full year of development.
The first Pāli Platform was written in Java 8. The main
reasons for using Java are: First, Java executables can run
in all major operating systems, as its motto says, “Write once,
run anywhere.” And second, Apache Lucene, an excellent IR
system, is written in Java.
There was a limitation at the time. Because I often traveled
from place to place, mostly by foot. I had to keep my belongings
minimal, able to carry along in a bag or two. From that limi-
tation, the first release of Pāli Platform was totally developed
in Raspberry Pi 3 Model B, a credit-card-sized computer, with
5-inch LCD display. As a result, it forced me to use Swing, an
old, a little outdated, Java user interface (UI) toolkit.2
After I made use of Pāli Platform 1 to write two Pāli learn-
ing books. I have rewritten the program again from the ground
up, using Java 11 and JavaFX UI (around 10% of code was
reused). Doing the same thing twice can get us some insight.
The result of my effort this time is really rewarding. I just love
it, and I hope students of Pāli would feel in the same way, and
use the program as their study/research platform.
1.2. Why I take Pāli seriously?

Let me use this section to explain that why I spent my time
and effort to do my Pāli related projects. As a matter of fact,
I neither work in the software industry nor academia. So, I
am not interested in producing programs to sell, nor pursu-
ing academic positions. I just live my life peacefully, mostly
in solitude (and a kind of destitution). That makes me have
sity, and finished it in the following semester.
2 In fact, JavaFX was available. But for ARM devices, it was incomplete at
the time. Another reason was the Swing UI is really well-documented, easy
to learn and use.
4
1.2. Why I take Pāli seriously?
plenty of time to do what I think it is worth spending the day

to day hours. This means the projects is closely related to my
religious life, but not in the way Buddhists expect.
I do not think mastering Pāli will lead anybody to enlighten-
ment. Many liberated ones do not know a word of Pāli. Like-
wise, knowing everything stated in the Pāli canon cannot lead
anyone to liberation. The all Pāli related matters are about
intellectual enterprise, as well as political strategy. This may
raise a big issue of debate, but this is not a good place to go
into that. In my view, most of Pāli matters are power related
and about thought control3 , little to do with real liberation.
It is undeniable that the Pāli canon constitutes the bedrock
of all Theravāda traditions and cultures. Most cultural values
in Theravāda countries are derived from the canon. The only
key to ‘decipher’ the meaningful messages from the text is Pāli
language. The word ‘decipher’ is tricky. It sounds like there
is the only way to decode the messages. In other words, you
cannot read the text in whatever way you want. There is the
only way, and the right way is maintained by the tradition.
That is an illusion. The illusion that the tradition (read, those
who gets the power) tries to implant in the Buddhists’ mind.
I think, that is why learning Pāli in the traditional way is ex-
tremely (but unnecessarily) difficult. Those who can pass the
laborious process are counted as ‘authority.’ Then they can de-
termine what should be read from the canon. Outside of that
scope, it is regarded as unorthodox, or even blasphemous. By
such a system of learning, the ability of reasoning is impaired,
if not destroyed altogether.
From my engineering background, I was trained to think
that using a better tool brings a better result with less effort.
With a good learning tool, now those who have some effort (not
that much as the traditional students) and perseverance can
become ‘authority’ themselves. They can investigate why the
tradition gives certain explanations. And they can use reasons
whether to believe or embrace those positions or not.
That can emancipate Buddhists from the yoke of surrepti-
3 Some may argue it is also about knowledge. That is partly true. The main
purpose of such knowledge is a kind of power, ability to control and maintain

the superior position. We have to understand it nonetheless, to guard our-
selves against that kind of manipulation. So, for me, understanding of Pāli
leads to knowledge in this manner.
5
tious manipulation of power through religion. That is what I

have been trying to do. I have built a good tool, introduced an
effective way to learn the language, and encouraged the use
of good reasoning to assess whether certain beliefs should be
taken or not.
That should be enough for this uneasy matter. I will never
raise the issue again. For serious readers, please read further
Part 1 of Pāli for New Learners, Book II.
1.3. Features so far

The program can do many things related to Pāli studies. Be-
fore we go into detail of these in our course, let us see the big
picture. I divide the groups of features into three:
Most of the text compiled in the Chaṭṭha

(1) Pāli text collection
Saṅgāyanā4 is here. In addition, the users can add their own
text to the collection, in both XML5 and plain text format,
known as the Extra. The text can be accessed easily by a tree
of grouped contents. I call this TOC Tree. Or a specific text
can be searched by various tools provided, from very simple to
mind-bogglingly complex.
(2) Essential grammatical toolsNew Pāli learners will find many

helpful resources here. They have indispensable Pāli dictio-
naries to consult with, a bit old, but still useful. They can
learn various Pāli scripts, including Roman, Devanagari, Sin-
halese, Khmer, Myanmar, and Thai. They can learn inflec-
tions of nouns and verbs, by typical paradigms, as well as by
experimenting with generic rules. The number generator can
produce Pāli numerals up to 6 digits. We also have a table of
Pāli roots as listed in Saddanīti Dhātumālā. Advanced stu-
dents can find learning Pāli prosody is easy and fun with a
powerful stanza analyzer. (We even have meter search in Sim-
ple Lister and Tokenizer to facilitate prosodic composition.)
4 The collection is known as Chaṭṭha Saṅgāyana CD (CSCD), distributed
by Vipassana Research Institute (VRI), tipitaka.org.

5 The format is conformed to the collection’s.
6
1.4. How to run the program
(3) Powerful search tools Searching is the primary goal of the

program. So, we can see various approaches for finding things
here. At basic level, texts in the collection can be searched by
Document Finder, both by text name (as shown in TOC Tree)
and by text content (brute full text search). For a more ad-
vanced search, Lucene Finder can be used with its full capac-
ity. The users can build Lucene’s index with options adjusted
as needed. For a custom index builder, the users can use To-
kenizer instead, the IR module developed by the programmer.
This allows the users can add any number of documents, in-
cluding the Extra, to the index. The search function of Tok-
enizer may be not so powerful as that of Lucene, but it is good
enough for a specific search and it can list all terms in the in-
dex (not implemented in Lucene Finder). Or, if the users just
want to see the list of all terms in the collection. The pre-
built list can be found in Simple Lister. A specific term can be
searched here.
(4) Unique and powerful accessories The program also incorpo-

rates a number of general tools, such as Pāli Text Editor (with
several helpful functions), Pāli Text Reader with Sentence Man-
ager (with ability to help text reading and to add translations),
and batch script transformer (to convert a bunch of text files
between Roman and other scripts). Furthermore, I introduce
new ways of entering Roman Pāli character, which are more
comfortable than the old methods.

Now we come to the technical instruction. If you have not got
the program yet, go to paliplatform.blogspot.com6 and down-
load it from the link provided. The file you get looks something
like this: PaliPlatform2.0-?.zip
The ? stands for its subversion, e.g. RC1 (Release Candidate
1) or whatever. The file is packed in zip format. There are
two options: if you use Windows, you should download the full
version with JRE bundled (around 150 MB in size); if you use
Linux or macOS (or Windows, for those who plan to install
6 This is the only entry point I maintain. It can be changed in the future.
7
or add JRE by themselves), you can choose to download the

lesser version (around 90 MB) instead. Any common unpacker
should be able to explode the file.
Once you unpack the program to a writable place, you will
see a folder named PaliPlatform2 with the structure depicted
in Figure 1.1 (What is really shown in your computer is likely
different, but the content is the same).
Figure 1.1.: PaliPlatform2’s folder structure
Now you are ready to run the program. The principle is sim-
ple: The Java Virtual Machine (JVM), version 11 or newer,
will pick up the executable and run it. The JVM is a part of
Java Development Kit (JDK). For the users, you do not need
the whole JDK, just the part called Java Runtime Environ-
ment (JRE) is enough to run the program. The process can
be different from platform to platform as shown below. The
8
required computer spec is low, 2-GB RAM is safe, but more is

better. I even use an old 32-bit laptop in the whole process of
development.
1.4.1. Windows
Even though my developing environment is Linux, I make it
easy to run in Windows (7 or newer). For Windows, I propose
three use contexts:
(1) Using the bundled JRE This is the easiest way. The required
JRE is already shipped with the program under jre folder.
The only action you need is double-clicking PaliPlatform2.exe
to make the program start. That is all. You also can make a
desktop shortcut from this exe file.
STICKY-NOTE Note: To ensure that every machine can run the

program, I bundled only 32-bit JRE. For those who have
a 64-bit computer, you may need to replace the exist-
ing JRE with 64-bit version (see download links below),
then select PaliPlatform2-64.exe instead.
Replacing the existing JRE can also be done in the
case that you want to use your own JRE or to update it
to the latest version. To do so, delete the existing jre
folder, unpack the JRE in the program’s root, rename it
to jre. Then run PaliPlatform2.exe or PaliPlatform2-
64.exe for 64-bit JRE.
There is a good chance that your

(2) Using locally installed JRE
computer has a JRE installed. However, not every JRE is
usable. You need a JRE with JavaFX included. There are
some vendors who provide this (see links below). When you
download the JRE, make sure you get a full JRE or JRE with
JavaFX suitable to your machine. Once the JRE is installed
into the system, you can run the program by double-clicking
PaliPlatform.jar, or open it (by a right-click) with OpenJDK
platform binary or the like. You may test this first if your
existing JRE is usable.
9
(3) Using manual method This approach does not use the exe file
but use the bundled JRE, and you have to do it by hand. First
open a console terminal (command prompt)7 at the program’s
root, then enter this:
» jre\bin\java -p .;lib -m paliplatform/paliplatform.PaliPlatform
If you have a full JRE installed, you can use java directly,
hence:
» java -p .;lib -m paliplatform/paliplatform.PaliPlatform
1.4.2. GNU/Linux
This is my beloved operating system (OS). Although the num-
ber of Linux’s users is less than Windows and macOS, Linux
is very important OS nowadays. It liberates us. For Linux’s
users, mostly power users, it looks trivial to tell what to do to
run a Java program. So, I will leave out some details.
(1) Using an installed JREThere are two ways to install a JRE

into your system. First, you can install Java packages from
your Linux distro’s repository. For example, if you use Debian-
based OS, enter this command:8
» apt-get install openjdk-11-jre openjfx
Second, you can install JRE with JavaFX from other providers
(see download links below). Once, you have JRE and JavaFX
in your system, open a console at the program’s directory, and
type this launcher script:
» ./run.sh
If a separate JavaFX is used, you have to edit the launcher
script accordingly.
(2) Using a custom JRE If you do not want to install Java into
your system, you can just download an archive (zip or gz) ver-
sion and unpack it to the program’s directory. Rename it to
jrefx. Then type this launcher script:
» ./runfx.sh
7 In Windows 10, you can open a terminal at any place in File Explorer
by using the File menu. In Windows 7, I find it more difficult. You have to
search for command prompt or cmd and open it, or hit WINDOWS-R and enter cmd, then
make your way to the target directory by dir command.
8 If you already have a higher version of Java installed, just openjfx is
enough.
10
If you name the directory otherwise, editing the script is

needed. The script works only for full JRE with JavaFX in-
cluded.
(3) Using manual methodIf you prefer doing things by hand,

try this:
• If you have JRE with JavaFX included, type this at the
program’s root:
» jre_somewhere/bin/java -jar PaliPlatform2.jar &
• If you have JRE and JavaFX separately, use this instead:
» jre_somewhere/bin/java -p ".:lib:javafxdir/lib" ¬
-m paliplatform/paliplatform.PaliPlatform &
Notes: It is important to run these commands from the pro-
gram’s root directory. And it is advisable to set JAVA_HOME
and export it to PATH. Then, you just type “java ….”
In directory util, there are helper

(4) Creating desktop launcher
scripts to create a Linux desktop launcher for the program.
There are two scripts that make use of run.sh and runfx.sh.
For more information, please read the instruction there.
1.4.3. macOS
I have never owned or used any of Apple’s devices. For the

sake of testing, however, I spent many hours to have macOS
Catalina (10.15.7) installed in my Linux’s virtual machine.
Here are successful methods.
(1) Using an installed JREThis is the easiest way. First, you

have to download a full JRE (with JavaFX) installer from any
provider (see download links below), and install it in the sys-
tem.9 Then you can make the program run by double-clicking
PaliPlatform2.jar, or open it (by a right-click) with Jar Laun-
cher. Or, you can open a terminal in the program’s directory
and type this:
» ./run.sh
9 The installer packages can be either of .pkg or .dmg type.
11
(2) Using a custom JREIf you do not want to mess up with the
system, you can download an archive package instead and un-
pack it in the program’s directory. Rename it to jrefx, then
type this:
» ./runfx.sh
(3) Using manual method Since macOS is a kind of UNIX-like

OS, the Linux’s methods described above can be applied in
most cases. However, Linux desktop launcher does not work
in macOS. You have to create an alias of PaliPlatform2.jar
and move it to your Desktop.
1.5. When things go right
Once you succeed to make the program run in your computer.

You will see the main window as shown in Figure 1.2.
Figure 1.2.: Main window of Pāli Platform 2
You can check your runtime environment by choosing Help>

About in the menu, or clicking INFO-CIRCLE button. The system’s infor-
mation will show on the left of About window (see Figure 1.3).
12
1.6. When things go wrong
Figure 1.3.: Information of runtime environment
1.6. When things go wrong
Even though the program was well-tested, many bugs are still
waiting. They can show up when encountering unexpected
use cases. If the users find that the program behaves in an
unusual way or even crashes, please report the bugs.
To see error messages fired by the JVM, you have to run
the program with a console, i.e., you have to run the program
manually and leave the terminal opened. Error messages may
look unintelligible to the users, as shown in Figure 1.4, but
they indeed give useful information to trace the bugs’ causes.
When this occurs to you, please record the symptom and
save the message, then send it to me.10 I will find the causes
and fix the bugs.
10 jakratep at gmail dot com
13
Figure 1.4.: An example of the JVM’s error message
1.7. Download links

Here are some useful links related to the program’s execution.
BellSoft provides installers of JRE with

(1) BellSoft Liberica JDK
JavaFX for major platforms. This can be easy for new users
who just want to install what is needed. When downloading,
select JRE Full.
https://bell-sw.com/pages/downloads/#/java-11-lts
(2) Azul Zulu JDK Azul Systems also has JRE with JavaFX (32-
and 64-bit) for many platforms, but few have installers. I used
JRE from this vendor in the program’s bundle. When down-
loading, select JRE FX.
https://www.azul.com/downloads/?version=java-11-lts ¬
&package=jre-fx
(3) Eclipse Temurin JDK This provider does not provide JavaFX,
just JRE. This may be suitable for a competent user who knows
how to combine the two components together. (It is not that
difficult, particularly when you use Linux. See the manual
method described above.)
https://adoptium.net/temurin/releases/?version=11
14
1.7. Download links
(4) Gluon’s JavaFX This is the official site of JavaFX. You rarely
need this, except if you want the newest version of it (really
unnecessary for us). When downloading, select SDK binary
package, not jmods.
https://gluonhq.com/products/javafx
15
Basic operations and settings
2
Now, I suppose that the reader can run the program success-
fully and get the first screen. Unlike Pāli Platform 1 which
integrates all working areas in one window, this version uses
multiple windows approach. So, you will see many windows
doing various jobs. There are two kinds of window: single-
ton and multi-instance. The former has only one instance in
the program’s lifetime, while the later can have multiple in-
stances. Let us start with the main window.
2.1. Main window

The main window is a mandatory singleton. You only have one
main window when you run the program. You can start the
program multiple times, however, to create multiple running
instances. But, it is unnecessary, and not recommended, to do
so. Therefore, you are supposed to have only one main window.
There are three parts in the main window: the menu bar,
the main tool bar, and tabs of some working areas (see Figure
1.2). The first two parts are familiar to most computer users.
Every button in the tool bar has its counterpart in the menu.
They are selected to show here because they may be used a
lot. We will see the use of each command in the menu in due
course.
The third part is the big area below those two. There are
16
2.2. Common tool bar
fixed tabs of the most-used 5 modules: TOC Tree, Document

Finder, Simple Lister, Tokenizer, and Dictionaries. These tabs
are always there, no more no less, but you can reorder them as
you want. Each tab can be opened as new multiple windows
using the menu or the tool bar. When a new window is created,
it works separately from the main tab area. For referencing,
therefore, we will call these 5 tabs as the main TOC Tree, the
main Document Finder, and so on.
2.2. Common tool bar

Every working tab, also every opened window, has its own tool
bar that has a part in common (see Figure 2.1).
Switch between light and dark theme

Decrease font size
Reset to normal size
Increase font size
Change font
Save screen’s image
Save data as text
Copy data as text
Figure 2.1.: The common local tool bar
When you change theme by the tool bar, it changes only that
instance. If you click this button in the main tabs, the whole
main window is changed. This change is not persistent. If you
want permanent theme change, use menu Options>Global theme
instead. Other buttons are intuitively understandable, but
read more about font in its section below. Another common
button you can see here and there is help (Question-Circle) button. When
clicked, the related help of that part will show up.
17
2. Basic operations and settings
2.3. Fonts and problems

The users will not feel the significance of fonts and their prob-
lems until they read or edit a Pāli text. However, you should
know this before you find out yourself that things do not al-
ways go as you expect.
Fonts used in the program have two kinds: embedded and
external. The former has nothing to do with the users. You
may be able to select some of them, but they cannot be changed
or removed. This includes DejaVu1 font family (Serif, Sans
Serif, and Mono space) and icon fonts. So, when applicable,
you can always choose either DejaVu Serif or DejaVu Sans or
DejaVu Sans Mono. These fonts should have no problem in
most running environment.
Problems mostly come from other scripts than Roman. That
is why external fonts is needed. The idea is simple: put your
font files (TTF) in folder fonts at the program’s root and restart
the program. You should add two files: one for regular face,
another for bold face. If only regular face is added, you will see
only regular text, no bold, in Pāli documents.
Not every font is usable, however. For displaying Pāli Ro-
man, you must have a Unicode font with Latin Extended Ad-
ditional block (not every Unicode font has it). I added Times
Ext Roman2 as an example.
For other scripts, I added Arundina3 font for Thai, and Noto4
fonts for the others. Here is a warning: Not all fonts work as
expected. I always have problems with Myanmar fonts. The
provided Noto Myanmar works fine with my current Linux but
not the former one. When I tested with Windows, the Myan-
mar font did not work in Windows 7, but worked without prob-
lem in Windows 10. Fonts of other scripts may look weird in
some machines.
Here are some solutions when problems related to external
fonts occur:
If your system has proper fonts

(1) Just delete the external fonts
installed, you can delete fonts provided in the font folder (or
1 https://dejavu-fonts.github.io
2 This font, copyright of Monotype Corporation, was shipped in CSCD.
3 https://github.com/tlwg/fonts-arundina
4 https://github.com/notofonts/noto-fonts
18
2.4. Pāli input
move them away). Then, the program will use a system font
instead as a fallback, currently set to Sans.
For example, some versions of Noto

(2) Try different font versions
fonts work in some systems, some do not, or behave oddly. Try
another version of them may help.
I found that Padauk, Myanmar3, and

(3) Use alternative fonts
MyMyanmar font work fine in Windows 7. KhmerOS fonts
also work in various contexts. LKLUG (Sinhala) font also works
well in most Linux.
2.4. Pāli input

Typing in Pāli characters is used throughout the program. So,
it is essential that the users have to learn the input method.
It is easy, but needs some familiarity. In text fields and text
areas that expect Pāli Roman characters, there are 3 modes of
input indicated by their symbol: Regular mode (aIa), Unused-
characters mode (xIā), and Composite mode (āIā). These three
modes can be switched circularly by clicking the symbol button
or pressing Ctrl-Space. Pāli input also has detailed settings
in Settings window, invoked by menu Options>Settings or COG
button in the main tool bar.
aIa Regular mode is the normal state of your computer. As
the symbol says, when you type ‘a’ you simply get ‘a’. You can
type in English or other languages supported by your system
in this mode.
xIā Unused-characters mode is a unique and innovative
method. This mode utilizes unused English characters for
some Pāli letters with diacritics. Here are the program’s de-
fault mapping: x = ā, X = ī, w = ṃ, W = ū, F = ṅ, f = ñ, q = ṭ,
Q = ḍ, z = ṇ, Z = ḷ. That is why the symbol says thus: you get ‘ā’
when you hit ‘x’, for example. This mode includes two special
keys for making a character under cursor uppercase (<) and
lowercase (>). All these keys can be set in Settings (see Figure
2.2).
āIā Composite mode is the most intuitive way to enter a
diacritic mark. Here are the program’s presets: ∼ = tilde,
19
Figure 2.2.: Settings for Pāli input
- = macron, ’ = dot above, . = dot below (see also Figure 2.2).

For example, when you need ‘ā’ you type ‘a’ followed by ‘-’ im-
mediately. The mapping can be set otherwise in the program’s
settings. This mode is a little more expensive than the previ-
ous one. You need two strokes rather than one, but you do not
need to remember the character mapping.
The users can set their preferred method as the default in
the program’s settings (see Figure 2.2).
STICKY-NOTE
Note: All user’s preferences, such as the
global theme, bookmarks, and those in Settings,
are saved in an external property file, named
PaliPlatform2.properties. This file may exist in
the user’s home directory or the program’s root or
somewhere else depending on how the program is
invoked. If you mess up with the settings, you can
return to the original state by just deleting the property
file and restart the program.
20
2.5. Minor concerns
2.5. Minor concerns
There are some miscellaneous points that are worth mention-

ing here, before you make a real use of the program.
(1) Drag-and-drop The program supports the drag-and-drop ac-

tion in various places. In some tasks, it is mandatory, such
as adding files into a Tokenizer window. In some, it is just a
convenient tool, for example, a file can be dragged from the
system’s explorer and drop into the program’s editor, or a text
portion can be dragged from a system’s editor and drop into
a Dictionaries window, or a text portion in the Dictionaries’
result can be dragged to somewhere else.
There are many possibilities of drag-and-drop. Some look
obvious. Some seem likely applicable, but it cannot be done
(because of a technical limitation). It will be too fussy to list
all of them here. Learning by using is better. If any drag-and-
drop action does not work, using copy-and-paste, if applicable,
can achieve the same goal in most cases.
(2) Context menus Context menus are menus that pop up when
you right-click (or left-click in some places) at a certain ob-
ject. These menus are context-dependent. You have to test
and learn by yourself. In many cases, context menus provide
more solid functions, than drag-and-drop does.
(3) Sortable tables There are a number of tabular results pro-

duced by the program. Most of those tables are sortable. It can
be done by clicking on the table header of the target columns.
The mode of sorting is changed circularly: ascending, descend-
ing, and unsorted. A triangle in the header will show up ac-
cordingly. Some tables, to which sorting makes no sense, can-
not be sorted, though. There are some things to keep in mind.
First, most tables are already sorted by a pre-selected crite-
rion. Using a custom sort can disrupt the preset order, in the
way that mode of sorting may be shown incorrectly. And sec-
ond, the sorting is done only on the existing data in the tables.
No new retrieval or calculation is applied.
21
(4) Tooltips When you are stuck in the using, try Question-Circle button
first. If the help is not available, try reading tooltips by hover-
ing the mouse over an unknown button. Tooltips can be seen
in other places as well, such as in a list or table that has trun-
cated data entries.
22
Part II.
Grammatical Tools
23
Dictionaries
3
We will start with grammatical tools, because they are not
complex and rather easy to use and understand. All of these
tools can be invoked by menu Grammar (Figure 3.1). Only Dic-
tionaries has its button (BOOK) in the main tool bar.
Figure 3.1.: Menu Grammar
Dictionaries is the most used tool. It occupies one tab in the

main window, and it can be opened as separate windows, as
many as you want. A Dictionaries window is shown in Figure
3.2.
There are four dictionaries you can search in this module.
(1) Concise Pāli-English Dictionary (CPED)1 This dic-
tionary holds a significant position in the program. It is the
1 By Ven. A. P. Buddhadatta, available online at https://www.budsas.org/
ebud/dict-pe/index.htm
24
Figure 3.2.: Dictionaries window
only one that is structured and used widely in various con-

texts. I also made a few corrections and added some entries.
(2) Concise English-Pāli Dictionary (CEPD)2 This may
be less used. Sometimes it can give you a word that you do not
know it in Pāli, but you know its close English words.
(3) The Pali Text Society’s Pali-English Dictionary
(PTSD)3 A bit old, this dictionary is still indispensable and
often gives some insight to the words we study.
(4) Dictionary of Pāli Proper Names (DPPN)4 You may
seldom use this one. It gives detailed information about Pāli
names, not directly the explanation of the words.
The inclusion of these can be set by their check boxes in the
tool bar. For persistent setting, consider using Settings win-
dow as shown in Figure 3.3.
2 By Ven. A. P. Buddhadatta, available online at https://www.budsas.org/
ebud/dict-ep/index.htm
3 By T. W. Rhys Davids and W. Stede (1921–25), available online at https:
//dsal.uchicago.edu/dictionaries/pali
4 By Gunapala Piyasena Malalasekera, available online at http://www.
palikanon.com/english/pali_names/dic_idx.html
25
3. Dictionaries
Figure 3.3.: Default dictionary selection
Searching in Dictionaries can be of three use cases as fol-

lows:
1. Simple search
Simple search is fast and simple. You enter some characters,

the relevant result will listed immediately, and the topmost
entry is shown in the result area. Figure 3.2 shows the result
of a simple search.
2. Wildcard search
Sometimes you have a vague idea what the word looks like.
You can use wildcard mode by check at Use */?. In this mode
you can use ? to stand for any one character and * to stand for
any characters (including none). Figure 3.4 shows the result
of searching ‘*mutt?’. Remember that in wildcard mode the
result does not come up immediately. You have to press Search
button or hit the Enter key.
As shown in the picture, * can mean zero character, and
? always represents one character. So, we also see mutta,
muttā, and mutti in the result list. In the result of CPED,
if the selected term is declinable (either noun or adj.), we will
see Show declension . When we click on this box-button, a
declension window of that term will be opened.
26
Figure 3.4.: Dictionaries’ wildcard search
3. Meaning search
Yet sometimes you have no idea what the word should be. In
this case, searching in dictionaries’ meanings can help. Fig-
ure 3.5 shows the result of searching ‘emancipate.’ In this
mode, using wildcards is not allowed, and do not forget to press
Search or hit Enter to submit the query.
There is a technical limitation in highlighting the search
result, when you search in meanings. Also, you cannot search
further in the display result. To do this, you have to open the
result in a text editor, and do the search there. This can be
done by hitting PENCIL-ALT button in the tool bar. Then the result will
be opened in the program’s editor. Then the users can use
search function in the editor, as shown in Figure 3.6.
27
3. Dictionaries
Figure 3.5.: Dictionaries’ meaning search
Figure 3.6.: Dictionaries’ result opened in an editor
28
4
Letters
With a good tool, learning Pāli alphabet can be an enjoyable

experience. The Letters window can be opened by selecting
menu Grammar>Letters. In its tool bar, there are additional
buttons:
E/P Switch between English and Pāli technical terms

Broom Clear highlights and selections
Language Select a script language
Keyboard Open/close typing test area
The display of script languages other than Roman, i.e., De-

vanagari, Khmer, Myanmar, Sinhala, and Thai, depends on
external fonts loaded when the program started. If any prob-
lem occurs, read the instruction in Section 2.3. Other oper-
ations look obvious. So, I leave the detail to the user to find
out.
29
4. Letters
Figure 4.1.: Letters window
30
Declension table
5
This is one of the most useful tool for new learners. Tradi-
tional students have to remember many of these tables. While
remembering some key paradigms is still important in learn-
ing process, this tool can enhance your ability to search and
experiment in just a few clicks. The Declension window can
be opened by menu Grammar>Declension table. The results of
Declension table can be grouped into three: nouns and adjec-
tives, pronouns, and numbers.
5.1. Pronouns
If you are new to this window, I suggest you select Pronouns
first because the pronoun list is finite. You will see some-
thing like Figure 5.1. The result of declension tables will show
one gender at a time. If there are multiple genders to be de-
clined, you can choose whether m (masculine), f (feminine), or
nt (neuter). Short meaning of each pronoun can also be added
when this option is selected.
For new students, the expansion of abbreviations used in
the table, as shown in Table 5.1, may do some help.
An amazing feature of this tool is that you can check the de-
clined terms against term list in the collection. When LIST but-
ton is hit, you will see the right pane opened (see Figure 5.2).
This may take a little time in the first load. In the list, terms
31
5. Declension table
Figure 5.1.: Declension table of a pronoun
Table 5.1.: Expansion of case abbreviations
1 paṭhamā nominative case

2 dutiyā accusative case
3 tatiyā instrumental case
4 catutthī dative case
5 pañcamī ablative case
6 chaṭṭhī genitive case
7 sattamī locative case
ā ālapana vocative case
in the table will be sorted by their frequency. Then, you can

know which forms are actually used in the texts. Remember
that one declined form of a Pāli word can come from different
bases, for example taṃ and te can come from tumha or ta.
There are two options for the source of term list used here
(selected by Check-double button above the list): The first is from the over-
all list, every terms in the whole collection. And the second is
from the list produced by the main Tokenizer, the Tokenizer
tab in the main window. If you have not yet used the Tokenizer,
the list should be empty by this option.
32
5.2. Nouns/Adjectives
Figure 5.2.: Declension of tumha against term list
5.2. Nouns/Adjectives
Nouns and adjectives in Pāli undergo the same inflectional
rules, and sometimes a word takes dual position. So, I group
them together. Since terms in this group are numerous, you
have to select what you want by entering some query. Only
simple search is available here. The source of terms is the Con-
cise Pāli-English Dictionary (CPED). An example of noun’s de-
clension is shown in Figure 5.3.
Figure 5.3.: Declension of vimutti against term list
Results of an adjective’s declension look exactly the same,

but adjectives can decline into all three genders.
33
5. Declension table
What about the Compute button then? It has two uses. First,
when a term has both regular and irregular declensions. And
second, just when you want to play around with experiments.
I will give an example of the first case. For the second, the
users can do by themselves. There are many words that use
irregular paradigms, and I put a lot of effort to incorporate
them here seamlessly.1
A term I give here is sara. When it means ‘pond’ it declines
irregularly as mano-gaṇa (m. only). This is what is shown
when you enter ‘sara’ at the search box. But when it means
‘sound’ or ‘arrow’ it declines like normal nouns (m. and nt.). To
reach the latter case you have to hit Compute after you enter
‘sara’. See the comparison in Figure 5.4.
For those who want to experiment with generic paradigms,
you simply input a word with a proper ending and hit Compute.
The proper endings are a, i, ī, u, ū for masculine words, ā, i, ī,
u, ū for feminine words, and a, i, u for neuter words. There are
also two exceptions which use their special paradigms: -ant
and -ar.
5.3. Numbers
Numeral declension is one of the most difficult parts of the pro-
gram to implement.2 All output in this part uses calculation.
That means you can enter any number up to 6 digits, and it
will be converted to its Pāli numeral3 with declension tables,
both cardinal and ordinal. To help the beginners, however,
there are pre-built lists that can be easily selected. Figure 5.5
shows the result of 100250.
I will skip other details because everything is self-explained
and it is better to learn by playing.
1 If you find any irregular word that declines wrongly, according to the tra-
ditional textbooks, please report me right away.

2 This part is not long but complicated. Other difficult and quite big mod-
ules are Tokenizer, Pāli Text Reader & Sentence Manager, and the viewer of
Pāli documents including script transformers.
3 The result list is not exhaustive, because one complex number can be ren-
dered in many ways. The results here show only some compound units that
can be calculated by the computer. Try 150, 250, 350, and so on; you will see
why the conversion is so difficult.
34
5.3. Numbers
(a) Declined as mano-gaṇa
(b) Computed declension
Figure 5.4.: Declension of sara
Figure 5.5.: Declension of 100250
35
Verbs
6
Verbs listed here come from CPED. Fortunately, the dictio-
nary has already provided us the composition of verbs and
their related forms. New students can learn a lot just by going
through these lists.
We have only two main options: to see main (canonical) verb
form (singular, present, third-person, active), typically those
ending with -ti, and the rest. In main verbs, you can also filter
the list by the verbs themselves, or their root and prefix, or
their ending component (paccaya)1 , or their meaning. The use
is straightforward. Figure 6.1 shows a search result.
Figure 6.1.: Result of main verbs when search ‘vim’ and select
vimuccati
1 The name of roots and paccayas can be slightly different from Kac-
cāyana/Saddanīti’s convention.
36
In other verb forms, you have to choose the group you want
(see Figure 6.2). The names of verb forms here come directly
from the dictionary. So, some name may look unfamiliar, for
example, Potential Participle is called commonly by other text-
books as Future Passive Participle.
Figure 6.2.: Options of other verb forms
37
Conjugation table
7
Rendering verb forms is a formidable task for new students be-
cause there are many things to remember, as well as to know
how to put them together. Having a good tool that can provide
a good number of examples can speed up the learning process.
Unlike a declension table that can be created upon any declin-
able word, a conjugation table cannot be created for any root
because of the lack of complete uniformity.
With this variety of verb formation, we can do, at best, only
by providing some typical verb samples, as shown in the left-
sided list. And only a handful of verbs that have wide range of
forms covering most of tenses and moods, for example, pacati
(paca), gacchati (gamu), and bhavati (bhū). That is why the
program starts with pacati.1
There are two main modes in this module: main and deriva-
tion. Main verb formation is about ākhyāta. So, conjugation
makes sense only for main verb forms. Derivative verbs have
no conjugation, because they undergo different rules resulting
in different verb forms. Most of derivative verbs are declin-
able. So, we will see declension tables as a result of derivation
mode.
Like declension tables, we can check the words rendered
against an internal term list by opening the right pane (LIST
1 Rendering paca is straightforward and clean, having no weird variation.
It may be the most used example for teaching verb formation.
38
button). Opening the right pane can make rendering a new
verb sluggish, so use it when necessary.
Rendering a conjugation table has additional two options (Check-double
button): (1) whether Attanopada (middle voice) will be shown,
(2) whether Augment (a-)2 will be added. Figure 7.1 shows
aorist forms of pacati in full options.
Figure 7.1.: Aorist forms of pacati
To help new learners, I add Table 7.1 to show technical terms

used by Pāli textbooks, as abbreviated in the result area. How-
ever, for those who are new to Pāli grammar and not familiar
with verb formation, just playing around this tool can do little
help. So, you should read Pāli for New Learners, Book I, to
understand the main idea about Pāli verb formation first.
In Derivation mode, major derivative verb forms can be stud-
ied here in various combinations. The result is shown as de-
clension tables, except for tvā verbs. Figure 7.2 show mascu-
line ta forms (past participle) of pacati. Derivative verbs have
substantial details in Pāli grammar, so please see further in
the book mentioned.
Like Declension table window, we can experiment with any
verb stem here (see the option in Check-double button). In experiment
mode, the result will be rendered against the generic rules
of the verb form selected. This will be useful when the right
pane is also opened. In this way, the users can check whether
2 Augment is added only to some of Imperfect verbs and all of Aorist, and
Conditional verbs. The augment has no specific meaning, so it is a redundant

part. However, it is fashionable to be used with some verbs.
39
7. Conjugation table
Table 7.1.: Tenses and moods
English Pāli
Present tense Vattamānā
Imperative mood Pañcamī
Optative mood Sattamī
Perfect tense Parokkhā
Imperfect tense Hiyyattanī
Aorist tense Ajjatanī
Future tense Bhavissanti
Conditional mood Kālātipatti
Figure 7.2.: Masculine ta forms of pacati
a particular stem has its use in the texts.
40
Roots
8
In the traditional way of learning, to know a verb means to
know its root. So, it is really important to identify a root when
we encounter a verb. Learning Pāli roots, unfortunately, is not
an easy task because different schools name roots differently.
Furthermore, some roots can behave strangely in certain con-
texts. So, in modern way of learning Pāli, knowing roots is
less pressing. But knowing them still brings great benefits.
Roots listed here come from the compilation of Ven. U Silana-
nda in Pali Roots in Saddanīti Dhātu-Mālā compared with
Pāṇinīya-Dhātupāṭha.1 So, the roots are familiar to those who
follow Kaccāyana/Saddanīti school.
Things the users can do with the roots are: you can select
only a specific verb group, you can search in the roots’ name, in
Pāli meaning, and in English meaning. Figure 8.1 show Roots
window with an option opened.
In some entries, you can see an asterisk (*) mark in the root’s
name. In some, you can also see double asterisks (**) in the
Pāli meaning. These marks tell us that there is a comment in
the entry. You can see the comment by clicking the row in the
table. Comments are directly taken from the original work.
So, some of them can be difficult to understand exactly.2
1 Edited by U Nandisena, available at https://archive.org/details/
ThePaliRootsInSaddaniti
2 Smith mentioned by comments is the publication of Saddanīti edited by
H. Smith (1928–66), 6 volumes. In this context it means the second volume
41
8. Roots
Figure 8.1.: Roots window
The best way to study these roots, I suggest, is to read Sad-

danīti Dhātumālā (available in the collection under Byākaraṇa
gantha-saṅgaho in the Añña group) directly on the root you
are interested in. A quick way to access to the desired point
is to search the root and its Pāli meaning in that text group.
Learn more about search in its part below.
(Dhātumālā) of that work. The 5 volumes of the first edition can be found at
https://archive.org/details/SaddanitiAggavamsasPaliGrammar01 (to 05).
42
Prosody
9
Pāli prosody, or chanda, is a big, difficult topic to learn. I am
not keen on this subject, nor I am a poetry enthusiast. I see it
from an engineering point of view that if we have a good tool
we can learn this hard topic easier, or even with fun. There are
several things to learn before you can use this tool effectively.
When it is first opened, it shows just a list of prosodic types.
When you click on a row, you just see a bizarre formula at the
bottom border (see Figure 9.1).
Figure 9.1.: Prosody window
43
9. Prosody
For I have never mentioned Pāli prosody in my previous

books, I will do it in this chapter just enough to understand
how the program works. The main, in fact only, work that we
refer to is Vuttodaya.1
9.1. A survival introduction to Pāli prosody

As a subject, prosody is “The study of the metrical structure
of verse.”2 For our concern, it is the way to put words into
an organized structure ruled by groups of syllables. At the
fundamental level, a Pāli syllable can be either ‘light’ (lahu)
or ‘heavy’ (garu). To measure this ‘weight’ we consider only
vowels in a word.
Lahu is short open vowels When a, i, or u sits by itself alone or

follows a consonant, and not followed by a double-consonant
or ṃ, it is counted as ‘light.’ Lahu weighs 1 unit (measure).
So, we use symbol 1 and l here.3
Garu is everything else When you can identify lahu, garu is

simply the rest of that. Garu weighs 2 units. We use symbol 2
and g here.4
1 The author is Saṅgharakkhita of Udumbaragiri, around the 12th-13th
century AD, in Sri Lanka. In the collection, it is Vuttodayapāṭha, under

Byākaraṇa gantha-saṅgaho. The text is so terse that we can make little sense
out of it. Fortunately, I have Vuttodayamañjarī by Gandhasārābhivaṃsa,
printed by Mahachulalongkorn Buddhist University, Nakhon Pathom, Thai-
land (2001). This commentary on Vuttodaya is written in Thai. It is com-
prehensive and well-organized. I think this is what a commentary should
be. Writing new commentaries with modern languages is preferable nowa-
days because we can make them very clear to the readers, as exemplified by
this work. If you cannot read them, certainly someone can. Writing a new
text with Pāli makes things worse, and does little help to the textual under-
standing. For English readers, I find a translation by Ānandajoti Bhikkhu
useful, but not beginner-friendly, see https://www.ancient-buddhist-texts.
net/Textual-Studies/Vuttodaya/index.htm.
2 The American Heritage Dictionary, https://www.ahdictionary.com/
word/search.html?q=prosody
3 Many textbooks on the subject use ` for lahu and – for garu, or vice versa
(see Vut.7). I find this very confusing, so I avoid using these symbols.
4 The symbol 2 has double meaning. It means garu in the analyzed result
(as we shall see shortly). And if the symbol appears in a formula, it means
any combination that give 2 measures long, i.e., both ll (1+1) and g (2) can be
44
9.2. Two types of prosodic patterns
Let us see an example to make sure we get the same pic-

ture. In fact, I make a tool that can calculate the meter of any
text. It is a part of the program’s text editor. A result of meter
calculation is shown in Figure 9.2. For more information, see
Chapter 16.
Figure 9.2.: Meter calculation in the program’s editor
In the result, lahu and garu are denoted by 1 and 2 respec-

tively. First, buddhaṃ gets 22 because the u is followed by
ddh (double consonants), and the a is followed by ṃ. Second,
saraṇaṃ gets 112 because the first two a are an open short
vowel, whereas the last one is followed by ṃ. And third, gac-
chāmi gets 221 because the a is followed by cch (double conso-
nants), the ā is a long vowel, and the i is an open short vowel.5
Now we can see that when we put words together by a con-
straint of lahu-garu patterns, we are composing a verse. The
whole business of Pāli prosody is to recognize patterns and to
put words into patterns.

Broadly speaking, there are two ways to put syllables into pat-
tern: (1) by counting the syllables, and (2) by counting the
weight or measure. Understanding the distinction between
the two is important, so let make it clear first. We can illus-
trate the two types of counting by using the example above in
counted as 2. This may make some confused, but I insist on using 1 and 2
because they look better and clearer.
5 Some may wonder “What about a short vowel followed by a single conso-
nant (which does not belong to the next syllable)?” Put it bluntly, there is no
such a thing. You may see tad here and there in Pāli books, but not in the
traditional textbooks. We will not see tad stands alone. It is taṃ with d sub-
stitution as a result of word-joining (Sandhi). So, that syllable may be either
of lahu or garu type depending on whether the d is doubled or not.
45
9. Prosody
Table 9.1.
Table 9.1.: Syllable vs. weight summation
Word Meter Syllable sum Weight sum

buddhaṃ 22 2 4
saraṇaṃ 112 3 4
gacchāmi 221 3 5
By those ways of counting, Vuttodaya categorizes prosodic

patterns into two, namely mattāvutti and vaññavutti. The
former uses weight-counting, mainly with 4-measure groups
(gaṇa). The latter uses syllable-counting, mainly with 3-syllable
groups. Table 9.2 summarizes the mattāvutti groups used in
the program’s prosodic formulas, whereas Table 9.3 summa-
rizes the vaññavutti groups.
Table 9.2.: Meter groups used in mattāvutti
Symbol Group Notes

n 1111 Na gaṇa
s 112 Sa gaṇa
j 121 Ja gaṇa
b 211 Bha gaṇa
m 22 Ma gaṇa
4 any any of the above five
6 any any of the above five plus g or ll
8 any any combination of the above five
l 1 lahu
g 2 garu
2 any g (2) or ll (11)
In the tables, we can see that in mattāvutti there are 5 main

groups (gaṇa), all have 4-unit weight but variable syllables.
The names of gaṇa are traditional, but I use lowercase here
to distinguish them from vaññavutti types. In the latter case,
there are 8 main groups, all have 3 syllables but variable weight.
46
Some names are common, some are new. Note carefully on Na

gaṇa and Ma gaṇa, which share the name but not the content.
I use uppercase in the latter. The 14-lahu group (L) is used
only in the program, not in any textbooks. Some of mattāvutti
verse types use the symbols of vaññavutti in their formula.
For example, Vetālīya (Vut.29) uses R (212), Opacchandasaka
(Vut.30) uses R and Y (122), and Acaladhiti (Vut.38) uses L.6
Table 9.3.: Meter groups used in vaññavutti
Symbol Group Notes

N 111 Na gaṇa
S 112 Sa gaṇa
J 121 Ja gaṇa
Y 122 Ya gaṇa
B 211 Bha gaṇa
R 212 Ra gaṇa
T 221 Ta gaṇa
M 222 Ma gaṇa
l 1 lahu
g 2 garu
L 14 × l used only in the program
As described above, it is possible that one passage can be

analyzed by both methods, resulting in different patterns. Let
us see the analysis of our example above in Table 9.4. There
are many possibilities how to put syllables into a group even
when using the same method, as shown in the table.7 This
variety gives us a number of predefined patterns presented in
Vuttodaya, as we shall see in the next sections.
Before we are going to see the verse list, let me explain some
symbols used in verse formulas. I translated the descriptive
rules into computable formulas. So, they look weird to non-
programmer users. But the formulas are not difficult to read
if you try to understand the symbols. Beside of meter-group
symbols described above, there are also special symbols, as
6 To make it clear, using symbols in the formulas has nothing to do with
the tradition. It is a convenient way for computation in programming.

7 8/13 = syllable sum/weight sum
47
9. Prosody
Table 9.4.: Syllable vs. weight grouping
Passage buddhaṃ saraṇaṃ gacchāmi

Raw meter 22 112 221 (8/13)
Mattāvutti grouping 1 22-112-22-1 = m-s-m-l
Mattāvutti grouping 2 2-211-22-21 = g-b-m-gl
Vaññavutti grouping 1 22-112-221 = gg-S-T
Vaññavutti grouping 2 221-122-21 = T-Y-gl
shown in Table 9.5.
Table 9.5.: Symbols in verse formulas
Symbol Meaning
Vertical bar (|) logical OR
Exclamation mark (!) logical NOT
Semicolon (;) line separator
Hyphen (-) mere separator
We will see how logical OR and NOT work when I put a real
formula into explanation. The line separator (;), if present,
tells us that the pattern needs 2 lines of verse to make the
analysis complete. If the input passage is not long enough, or
does not have 2 lines as required, incomplete will be shown
in the status bar. But, if the analyzed text has more syllables
than the formula requires, ‘over-required’ will be shown in-
stead. Hyphen (-) is just a separator, having no meaning. It
only makes formulas easier to read.
9.3. Verse types of mattāvutti

Vuttodaya starts with definitions and general rules, then verse
types of mattāvutti are listed. We can divide further into 4
subgroups, namely, Ariyā, Gīti, Vetālīya, and Mattāsamaka.
All verse types (gāthā) in each subgroup are listed in Table
9.6. Not all of these can be analyzed by the program, as noted
48
below.
Table 9.6.: Verse types of mattāvutti

ID Name WS8 Parent group Ref.
0 Ariyā 57 Ariyā Vut.17
– Paṭhayā 57 Ariyā Vut.20
– Vipulā 57 Ariyā Vut.21
1 Capalā (full) 57 Ariyā Vut.22
2 Capalā (1st half) 30 Ariyā Vut.22
3 Capalā (2nd half) 27 Ariyā Vut.22
4 Mukhacapalā 57 Ariyā Vut.23
5 Jaghanacapalā 57 Ariyā Vut.24
6 Gīti (1st half of Ariyā) 30 Gīti Vut.25
7 Upagīti (2nd half of Ariyā) 27 Gīti Vut.26
8 Uggīti (Reversed Ariyā) 57 Gīti Vut.27
9 Ariyāgīti 32 Gīti Vut.28
10 Vetālīya 30 Vetālīya Vut.29
11 Opacchandasaka 34 Vetālīya Vut.30
12 Āpātalikā 30 Vetālīya Vut.31
13 Dakkhiṇantikā 30 Vetālīya Vut.32
14 Udiccavutti 30 Vetālīya Vut.33
15 Paccavutti 30 Vetālīya Vut.34
16 Pavattaka 30 Vetālīya Vut.35
17 Aparantikā (1) 32 Vetālīya Vut.36
18 Aparantikā (2) 32 Vetālīya Vut.36
19 Cāruhāsinī (1) 28 Vetālīya Vut.37
20 Cāruhāsinī (2) 28 Vetālīya Vut.37
21 Acaladhiti 16 Mattāsamaka Vut.38
22 Mattāsamaka 16 Mattāsamaka Vut.39
23 Visiloka 16 Mattāsamaka Vut.40
24 Vānavāsikā 16 Mattāsamaka Vut.41
25 Citrā 16 Mattāsamaka Vut.42
26 Upacitrā 16 Mattāsamaka Vut.43
– Pādākulaka Mattāsamaka Vut.44
Verse types in this category have complicated formulas. So,

let us go slowly. Normally, one stanza consists of 4 feet (quar-
9 Weight Sum, the total weight required
49
9. Prosody
ter, pāda) or 2 lines. Some verse types need 2-line pattern,

some need only 1-line to complete the analysis. This sounds a
bit tricky. Normally, we do not compose a single-line verse.
At least, we need two lines. However, many patterns have
repeated formula for the first and second line. As a result,
checking only one line for such patterns is enough. A sugges-
tion here is “Always analyze two lines of verse.” Let us see the
metrical pattern of Ariyā:
!j-4-!j-4-!j-j|n-!j-g; (30)
!j-4-!j-4-!j-l-!j-g (27)
The pattern needs 2 lines (see ;). The first line is read
NOT j, 4, NOT j, 4, NOT j, j OR n, NOT j, g. The second
is read NOT j, 4, NOT j, 4, NOT j, l, NOT j, g. As shown
in Table 9.2, 4 means any of meter groups is applicable, NOT j
means any meter group but j, and j OR n means either j or n
is valid.
Let us see a real example. This stanza is the first one of
Ganthārambhakathā in Dīghanikāya-aṭṭhakathā.
Karuṇāsītalahadayaṃ, paññāpajjotavihatamohatamaṃ;
112-211-112, 22-22-1111-211-2;
s-b-s, m-m-n-b-g;
Sanarāmaralokagaruṃ, vande sugataṃ gativimuttaṃ.
112-112-112, 22 112 1-112-2.
s-s-s, m s l-s-g.
If you carefully analyze the verse as shown above, you will

see that it conforms to the pattern described perfectly. To ana-
lyze the verse by the program, first open the text mentioned by
TOC Tree (see Chapter 10), select the stanza and copy. Then
open Prosody window and hit Analyze button. The most rel-
evant results will show at the top of the table. The analyzed
result against Ariyā pattern is shown in Figure 9.3.
In the result table produced by the program, Score is the
calculation of percentage of matching between the text and the
formula. So, 1.0 means 100% matched. There are possibilities
that we get multiple 1.0 results. But all of them are not always
‘correct’ answers. Some portion of text may be totally matched
50
Figure 9.3.: Analysis of Ariyā verse type
against the formula, but the text is over-required as shown

in Figure 9.4.
Another noteworthy point, even though a result is not 100%
matched, it can be a correct answer. Most often cases are the
unmatched last syllable of each line: it can be g despite the
required l, or vice versa. And as you may find out yourselves
later, exactly matched verses in the real text are quite rare,
except newly composed ones that are crafted in that way. The
real old verses are recalcitrant to the rules somehow. That is
why many exceptions are mentioned in detailed textbooks on
prosody.10
Now you can see how the program is really helpful in metri-
cal analysis. Doing it by hand is tedious, laborious, and error-
prone. Computer has its limitation, however. There are cer-
tain things unable to be analyzed, at least in a simple way, by
the program. Metrical pause (caesura), or yati in Pāli, is one
10 An attempt to shoehorn a verse into a recognized pattern creates excep-
tions, or extra considerations, to the rules. I do not use that approach in the
program. We just find the closest candidates and regard the verse as it is.
I think this is the normal way how poetry is composed. To achieve certain
artistic result, breaking some rules seems natural. Of course, there can be
cases of half-baked poetry that just ignores or plays with rules. And, finally,
there can be many cases that rules are held so firmly that the meter trumps
the clarity of the words used—words are oddly changed to conform with rules.
51
9. Prosody
Figure 9.4.: Example of an over-required case
of those.
Generally, metrical pause has two kinds: First, the pause af-
ter each foot ends. This is common in recitation and writing.
When you read a verse, you are supposed to pause at points to
make a sense of rhythm. And it is natural to pause at the end
of a foot, probably a small pause after the odd feet, a bigger
one after the first line (first even foot), and the biggest one af-
ter the second line (last even foot). The second kind of pause is
rarer, the pause inside the feet. Some verse types have rules
about pauses, particularly the Ariyā group. Pauses at the end-
ings is easy to see because punctuation marks can be used in
text. But in-between pauses cannot be shown in the program,
because this kind of pause makes sense only when the verse
is recited.11
Another limitation is the program cannot detect the spilled-
over effect: a word is cut into two and both parts are split
across feet. This aspect differentiates Vipulā (having a spilled-
over) from Paṭhayā (no spilled-over).12 As a result, the pro-
11 The rules concerning metrical pause are not easy to understand, so I skip
the explanation.
12 From Vut.21, “Yattha gaṇattaya mullaṅ-ghi, Yo’bhayatthā’dimo bhave vip-
ulā.” According to the formula, the word mullaṅghi has to be split so that the
-ghi part belongs to the latter foot.
52
9.4. Verse types of vaññavutti
gram can not tell the two verse types apart. They are grouped
together under Ariyā.13
It should be noted that Vetālīya (Vut.29) has a particular
rule: “No successive 6 lahus are allowed in the even feet.” This
means in its formula, 6-R-l-g-8-R-l-g, the 8 in the second
part does not allow neither 1111112 nor 2111111. This rule
has been already implemented in the latest program.
Finally in this category, Pādākulaka (Vut.44) is an arbitrary
combination of Mattāsamaka group. It has no fixed pattern to
detect, so it is left out from analysis (no ID).14

There are 3 main subgroups in this category: Samavutti (sym-
metry), Aḍḍhasamavutti (semi-symmetry), and Visamavutti
(asymmetry). The sense of ‘symmetry’ here means whether
each foot (quarter) of a verse is equal in length. It can be that
all 4 feet are equal (symmetry), or odd feet and even feet are
equal (semi-symmetry), or arbitrary (asymmetry). All verse
types in this category, plus some more, are listed in Table 9.7.
In Vut.13, twenty-six types of (symmetric) prosody (chanda)
are mentioned. That is to say, in theory, the metrical patterns
can run from 1 syllable to 26 syllables. Vuttodaya itself, how-
ever, lists only seventeen types: from 6-syllable (Gāyattī) to
22-syllable (Ākati) pattern (see other names in footnotes).
13 There is also a confusion of this. According to the text, in Vut.19 Ariyāsā-
mañña is mentioned, and Paṭhayā is mentioned in Vut.20, Vipulā in Vut.21.

So, it is supposed that Ariyāsāmañña and Paṭhayā are different kinds. But it
is not very clear how they differ to each other. In a way, we can interpret that
they are the same. By this concern, I drop Ariyāsāmañña from the list.
14 ID numbers used here have nothing to do with Vuttodaya. They are used
internally in the program. The users can use the numbers for sorting the
table, by clicking the column headers. To refer to a particular verse type,
using a reference to Vuttodaya, if any, is more reliable.
53
9. Prosody
Table 9.7.: Verse types of vaññavutti

ID Name SS15 Parent group Ref.
17
27 Tanumajjhā 6 Samavutti Vut.46
28 Kumāralalitā 718 Samavutti Vut.47
29 Citrapadā 819 Samavutti Vut.48
30 Vijjummālā 8 Samavutti Vut.49
31 Māṇavaka 8 Samavutti Vut.50
32 Samānikā 8 Samavutti Vut.51
33 Pamāṇikā 8 Samavutti Vut.52
34 Halamukhī 920 Samavutti Vut.53
35 Bhujagasususaṭā 9 Samavutti Vut.54
36 Suddhavirājita 1021 Samavutti Vut.55
37 Paṇava 10 Samavutti Vut.56
38 Rummavatī 10 Samavutti Vut.57
39 Mattā 10 Samavutti Vut.58
40 Campakamālā 10 Samavutti Vut.59
41 Manoramā 10 Samavutti Vut.60
42 Ubbhāsaka 10 Samavutti Vut.61
43 Upaṭṭhitā 10 Samavutti Vut.62
44 Indavajira 1122 Samavutti Vut.63
45 Upendavajira 11 Samavutti Vut.64
– Upjāti 11 Samavutti Vut.65
46 Sumukhī 11 Samavutti Vut.66
47 Dodhaka 11 Samavutti Vut.67
48 Sālinī 11 Samavutti Vut.68
49 Vātommī 11 Samavutti Vut.69
50 Sirī 11 Samavutti Vut.70
51 Rathoddhatā 11 Samavutti Vut.71
52 Svāgatā 11 Samavutti Vut.72
53 Bhaddikā 11 Samavutti Vut.73
54 Vaṃsaṭṭha 1223 Samavutti Vut.74
Continued on the next page…
16 Syllable Sum, the total syllables required
17 Gāyattī
18 Uṇhikā
19 Anuṭṭhubhā
20 Barahatī
21 Panti
22 Tiṭṭhubhā
23 Jagatī
54
Table 9.7: Verse types of vaññavutti (contd…)

55 Indavaṃsā 12 Samavutti Vut.75
56 Toṭaka 12 Samavutti Vut.76
57 Dutavilambita 12 Samavutti Vut.77
58 Puṭa 12 Samavutti Vut.78
59 Kusumavicittā 12 Samavutti Vut.79
60 Bhujaṅgappayāta 12 Samavutti Vut.80
61 Piyaṃvadā 12 Samavutti Vut.81
62 Lalitā 12 Samavutti Vut.82
63 Pamitakkharā 12 Samavutti Vut.83
64 Ujjalā 12 Samavutti Vut.84
65 Vessadevī 12 Samavutti Vut.85
66 Tāmarasa 12 Samavutti Vut.86
67 Kamalā 12 Samavutti Vut.87
68 Pahassiṇī 1324 Samavutti Vut.88
69 Rucirā 13 Samavutti Vut.89
70 Aparājitā 1425 Samavutti Vut.90
71 Paharaṇakalikā 14 Samavutti Vut.91
72 Vasantatilakā 14 Samavutti Vut.92
73 Sasikalā 1526 Samavutti Vut.93
74 Maṇiguṇanikara 15 Samavutti Vut.94
75 Mālinī 15 Samavutti Vut.95
76 Pabhaddaka 15 Samavutti Vut.96
77 Vāṇinī 1627 Samavutti Vut.97
78 Sikharaṇī 1728 Samavutti Vut.98
79 Hariṇī 17 Samavutti Vut.99
80 Mandakkantā 17 Samavutti Vut.100
81 Kusumitalatā 1829 Samavutti Vut.101
82 Meghavipphujjitā 1930 Samavutti Vut.102
83 Saddūlavikkīḷita 19 Samavutti Vut.103
84 Vutta 2031 Samavutti Vut.104
24 Atijagatī
25 Sakkarī
26 Atisakkarī
27 Aṭṭhi
28 Accaṭṭhi
29 Dhiti
30 Atidhiti
31 Kati
55
9. Prosody

85 Sandharā 2132 Samavutti Vut.105
86 Bhaddaka 2233 Samavutti Vut.106
87 Upacitta 22 Aḍḍhasamavutti Vut.107
88 Dutamajjhā 23 Aḍḍhasamavutti Vut.108
89 Vegavatī 21 Aḍḍhasamavutti Vut.109
90 Bhaddavirāja 21 Aḍḍhasamavutti Vut.110
91 Ketumatī 21 Aḍḍhasamavutti Vut.111
92 Ākhyānakī 22 Aḍḍhasamavutti Vut.112
93 Viparītākhyānakī 22 Aḍḍhasamavutti Vut.113
94 Hariṇaplutā 23 Aḍḍhasamavutti Vut.114
95 Aparavatta 23 Aḍḍhasamavutti Vut.115
96 Pupphitaggā 25 Aḍḍhasamavutti Vut.116
97 Yavamatī 25 Aḍḍhasamavutti Vut.117
98 Vatta 16 Visamavutti Vut.118
99 Pathyāvatta 16 Visamavutti Vut.119
100 Vīparītapathyāvatta 16 Visamavutti Vut.120
101 Capalāvatta 16 Visamavutti Vut.121
102 Piṅgalavipulā 16 Visamavutti Vut.122
103 Setavavipulā 16 Visamavutti Vut.123
104 Bhakāravipulā 16 Visamavutti Vut.124
105 Pathamabhakāravipulā 32 Visamavutti Vut.124
106 Tatiyabhakāravipulā 32 Visamavutti Vut.124
107 Rakāravipulā 16 Visamavutti Vut.125
108 Pathamarakāravipulā 32 Visamavutti Vut.125
109 Tatiyarakāravipulā 32 Visamavutti Vut.125
110 Nakāravipulā 16 Visamavutti Vut.126
111 Pathamanakāravipulā 32 Visamavutti Vut.126
112 Tatiyanakāravipulā 32 Visamavutti Vut.126
113 Takāravipulā 16 Visamavutti
114 Pathamatakāravipulā 32 Visamavutti
115 Tatiyatakāravipulā 32 Visamavutti
116 Makāravipulā 16 Visamavutti
117 Pathamamakāravipulā 32 Visamavutti
118 Tatiyamakāravipulā 32 Visamavutti
119 Sakāravipulā 16 Visamavutti
32 Pakati
33 Ākati
56

120 Pathamasakāravipulā 32 Visamavutti
121 Tatiyasakāravipulā 32 Visamavutti
122 Jakāravipulā 16 Visamavutti
123 Pathamajakāravipulā 32 Visamavutti
124 Tatiyajakāravipulā 32 Visamavutti
Vaññavutti verse types are relatively easier to understand

and analyze, but there is something to consider particularly
about symmetric types. Formulas given in Samavutti are only
for one foot, not one line or two feet (or full two lines or four
feet) as we have seen in the other group. As a result, you have
to edit the input passage by cutting one line of the verse into
two, otherwise ‘over-required’ will be the case. It is better to
see a real example, as shown in Figure 9.5.
Figure 9.5.: Analysis of Tanumajjhā verse type in edit mode
In the example, a line of Tanumajjhā (6-syllable T-Y pattern)

type is shown.34 In the text, we have stanzas in full form. If
34 The sample is taken from the 6th stanza of Mahāpaṇāmapāṭha, under
57
9. Prosody
you copy the whole line and analyze it, ‘over-required’ will
show. You have to cut the line by using edit mode (PEN-FANCY button),
then you will get the correct result. In semi-symmetry and
asymmetry groups, the formulas are given for one whole line,
so you do not have to do likewise in these groups.
There are some other notes the users should know. First,
Upajāti (Vut.65) is a mixed-up of Indavajira (Vut.63) and Up-
endavajira (Vut.64), or sometimes from other types. So, it has
no fixed pattern to detect, and left out from analysis (no ID).
Second, Sasikalā (Vut.93) and Maṇiguṇanikara (Vut.94) use
the same formula (14 lahus plus 1 garu) but the latter has two
pauses, after the 8th syllable and the last one. These two are
not distinguishable in the program, even if they are different.
And the last note, verse types ID 113 (Takāravipulā) to 124
(Tatiyajakāravipulā) have no references in Vuttodaya. These
types follow the logic of the previous Bhakāravipulā (Vut.124),
Rakāravipulā (Vut.125), and Nakāravipulā (Vut.126).35 I add
them here because they are easy to implement the analysis.
There are minor things I have not mentioned. The users
should play around and find out how things work. Once you
know this kind of tool exists, you can do research in Pāli prosody
easier than before. For metrical composition, this tool can be
your test bench. With edit mode mentioned above, you can
compose a short verse in real time. Furthermore, the program
also provides search by meter to help the users find a matched
word to a pettern required. We will talk about this in the re-
lated modules.
Buddha-vandanā gantha-saṅgaho in the Añña.

35 These are added in Vuttodayamañjarī by Ven. Gandhasārābhivaṃsa.
58
Part III.
Pāli Collection
59
Browsing and bookmarking
10
Before we learn how to access to a Pāli document in our col-
lection, it is better to be familiar with the collection first.
There are two kinds of documents used in the program: CSCD
and the Extra. The former are those bundled with the pro-
gram, whereas the latter should be empty at the first run.
The Chaṭṭha Saṅgāyana CD (CSCD)1 is the best and most
complete collection of Pāli literature nowadays.2 This is the
main corpus we use in Pāli studies. That is to say, Pāli Platform
is a one-stop package. You have everything essential for Pāli
learning in one place.
The Extra is a collection outside CSCD. It is just a directory
that can be set by the user (in General Settings).3 Two formats
are recognized as a document: XML (conformed to CSCD), and
plain text (with .txt extension). Once you have documents in
the Extra, you can open them in the program’s viewer, and you
can analyze them with Tokenizer. When documents are added
to the Extra while the program is running, UPLOAD button in TOC
Tree (see below) has to be pressed to make them visible.
A document in the collection can be accessed directly by TOC
1 Distributed by Vipassana Research Institute (VRI), tipitaka.org
2 In our application, Roman script is used as the base text.And its encoding
is changed from UTF-16 to UTF-8. All documents are structured in XML. All
files, including TOC files, are packed into one zip file, named romn_utf8.zip
in directory data/collection.
3 By default, it is set to data/collection/extra.
60
Tree, either its tab in the main window or a newly opened
TOC Tree window (using the Collection menu or the main
tool bar), as shown in Figure 10.1.
Figure 10.1.: TOC Tree window
The tree can be navigated by clicking its nodes. There are

four main groups of texts: Mūla (the main texts), Aṭṭhakathā
(commentaries), Ṭīkā (subcommentaries), and Añña (others).
For the first three groups, if you are familiar with the struc-
ture of the Pāli canon, it will be easy to find what you want
just by expanding the relevant nodes. Until you reach the text
level, you will get a context menu (see Figure 10.2). From this
menu, you can open the selected document either in the HTML
Viewer (see Chapter 12) or in bare text.
You have two options for opening a document in bare text,
with or without notes. The notes here are editorial insertion,
enclosed with square brackets. They are not a part of the text.
In certain situations, notes can disrupt the reading. So, they
should be filtered out when you need to analyze just the text.
Extracting bare text from a document in this way is useful
61
10. Browsing and bookmarking
Figure 10.2.: TOC Tree window at text level
when we prepare text for Pāli Text Reader (see Chapter 18).
From the context menu, you can also bookmark the docu-
ment as well as add it to Tokenizer. These two actions can
also be done by drag-and-drop. There are some buttons pro-
vided in the tool bar corresponding to the functions mentioned
above. Please check these by yourselves, as well as those un-
mentioned.
Adding documents to Tokenizer can be done with a whole
bunch of texts in a tree node by using GRIP-HORIZONTAL button in the tool bar,
because the context menu is available only at text level not the
higher levels.
In the main window, when BOOKMARK button is pressed, or it is se-
lected by the menu, Bookmarks window will show (see Figure
10.3). This window is a singleton. All documents you book-
mark in TOC Tree, or somewhere else, will appear here. Ac-
tions available, both by the context menu or the tool bar, are
similar to those of TOC Tree.
Bookmarks window is simple, so I will leave it to the users
62
Figure 10.3.: Bookmarks window
to find out what else it can do. There are some entries that are
preset in the bookmarks, two important documents difficult to
find by new learners. If you lose the preset, you can bring it
back by deleting the program’s property file (PaliPlatform2.
property) or move it away and restart the program.
63
Document Finder
11
Most new students of Pāli or Buddhism are not familiar with
the structure of the Pāli canon. Finding a text by navigat-
ing TOC Tree can be difficult. That is why Document Finder
comes in. With this tool, you can find relevant documents by
entering a query. If you know a certain name or a text portion,
you can find it more quickly than using TOC Tree. The tool is
a part of the main tabs, and you can open it as many as you
wish by clicking SEARCH button in the main tool bar, or selecting it
in the Collection menu.
What you should know first is there are two main kinds of
search here: heading search and content search. In head-
ing search, there are three fields available for searching: text
name, book name, and group name. What is counted as text,
or book, or group, is not exactly systematic. Those names come
from the organization of CSCD. They correspond with the en-
tries shown in TOC Tree. So, you may not find the document
you need by entering your familiar text name.
If heading search fails, you can resort to content search. In
this mode, full-text search will be applied, and the documents
containing the query will be listed.
The search modes mentioned can be selected in one place,
Check-double button. Figure 11.1 shows an attempt to find any document
in Dhammapada by searching in book name.
In heading search, an asterisk (*) can be used as a wildcard.
It can be used at any position (adding it to the last position is
64
Figure 11.1.: Searching Dhammapada in Document Finder
unnecessary). This can be helpful when you cannot remem-

ber certain part of the name. Figure 11.2 shows a use of the
wildcard in group name searching to find Visuddhimagga.
Figure 11.2.: Searching with a wildcard in Document Finder
Unlike heading search, in content search the query is always

treated as regular expression, and you have to hit the Enter
key to trigger the search, except when switching from other
modes. In this mode, the last column of the result shows the
number of hits in each document. Figure 11.3 shows the result
of an attempt to find Dhammacakkappavattanasutta with the
pattern ‘\b[Dd]hammacakkappa.*\b’ (see Chapter 20 for more
information).
Remember that this full-text search is case sensitive (know
your search query can narrow down the result significantly),
and brute (it searches directly in all files one by one, and you
have to wait until it finishes). Since complex patterns take
65
11. Document Finder
more time to process, it is better to use exact query terms

rather than meta-characters if you have those in mind. If the
search fails, consider using a more sophisticated tool, such as
Lucene Finder.
Figure 11.3.: Content searching in Document Finder
As shown in the picture, Document Finder has a context

menu similar to that of TOC Tree. So, you can access to the
document or do other actions in the same way.
66
Document viewer
12
After we know how to access to a document, in this chapter we
will learn about the viewer. As mentioned earlier, we can view
a document in two formats, XML or text. Here we focus only
on the viewing of XML documents. I call this tool Pāli HTML
Viewer.1
When a document is opened, either by a context menu or File-Alt
button in a tool bar, the viewer will show up. In its full form,
the viewer looks like Figure 12.1.
There are main three panes: (1) the center, always present,
displaying the text, (2) the right pane, opened by default, show-
ing the text’s information and related documents, and (3) the
left pane, hidden by default, used for navigation. The right
and left pane can be turned on and off by the three buttons in
the tool bar.
The information shown in the right pane is obvious. You can
also open further the related documents, if any, by using con-
text menu (right click). The related documents are the texts
that hierarchically related to the opened text. For example, if
you open a Mūla (main) text, the related texts in Aṭṭhakathā
(commentaries) will show. If you open a commentary, you will
also see links to the related texts in Ṭīkā (subcommentaries)
1 Technically speaking, the document is converted from XML to HTML us-
ing SGML transformation, then opened in the HTML viewer. The transforma-
tion has strict rules. So, non-compliant XML files cannot be viewed correctly
by the program.
67
12. Document viewer
Figure 12.1.: Document viewer in its full form
level.
The navigator in the left pane is quite useful. You can jump
to points in the document according to three criteria: heading
jump (HEADING), paragraph-number jump (¶), and stanza jump (Music).
The last point of jumping is saved. You can return there by
using the last jump button. To understand these you have to
experiment with various set of texts.
When you search a string by pressing Ctrl-F, the search
widget will show at the bottom of the window. There are three
options for searching: case sensitivity, whole word search, and
using regular expression. By default, the search is case-insen-
sitive and not whole-word. The options can be changed by Check-double
button. For a more advanced search, you can use regular ex-
pression (see more in Chapter 20).
When you right-click at a portion, or a selection, of text, a
context menu will show up, as shown at the center of the pic-
ture. This allows us to do certain operations upon the selected
text. The users should explore these by themselves. One ex-
planation, though, Send to Dictionaries means the selected
portion will be copied to the Dictionaries tab in the main win-
dow.
There are some buttons in the tool bar needed an explana-
68
tion. When STICKY-NOTE button is selected, the editorial notes embedded
in the text will show. The notes can be seen in blue text en-
closed with square brackets. When Hashtag button is selected, the
reference points to other publications will show.2
Figure 12.2.: Document viewer with Quick Dictionary
One useful feature of the viewer, is Quick Dictionary (using

CPED). When a word is selected, it will be searched in CPED
and the result is shown in a pop-up (see Figure 12.2). If the
exact word is not found, the nearest result is shown instead
with an asterisk (*) mark (see the picture).
Another exciting feature of the viewer is it can display text in
various scripts, other than Roman, namely Devanagari, Khmer,
Myanmar, Sinhala, and Thai. This can be done by Language but-
ton with an additional option, whether numbers are also con-
verted or not (0-9 button). An example in Myanmar script is
shown in Figure 12.3.3
There are some considerations and limitations concerning
script transformation. First, only Pāli characters are accept-
able. Converting Sanskrit characters produces unexpected re-
sult. Second, you can transform Roman script to any other
scripts, and vice versa. But you cannot convert a non-Roman
2 The publications are V (the VRI edition), P (the PTS edition), M (the
Myanmar edition), and T (the Thai edition). I have no idea what the exact
editions these refer to, and what the numbers are represented.
3 It is likely that you may encounter certain problem when displaying non-
Roman scripts. See the solution in Section 2.3
69
12. Document viewer
Figure 12.3.: Document viewer displaying in Myanmar script
to other non-Roman.
Figure 12.4.: General Settings of script transformation
Third, for to-Thai transformation, the users have an option

in the General Settings (see Figure 12.4) whether only the spe-
cial characters will be used or not. In Thai, ñ and ṭh have two
forms, with and without their lower part. The stripped forms
are displayed automatically in their specific contexts, other-
wise their full form (which looks less Pāli) will be used. If the
users want to use these special letters exclusively, they can
select the option. Remember that this can cause search prob-
lems in the native language. When converting from Thai to
70
Roman, this matter is not to be concerned because the pro-
gram can read both forms.
And finally also about the transformation of Thai vowel e
and o, as shown in Table 12.1, when Roman script is converted
into Thai, only on form is produced. On the other side, convert-
ing from Thai to Roman can tolerate the variation of input.
Table 12.1.: Transformation rules of Thai script
Input Output
tve
vho
tve
tve
vho
vho
71
13
Simple Lister
As we have learned so far, we can locate a document, open

it, read it with a help from Quick Dictionary, or transform it
to another script. Together with various grammar tools men-
tioned earlier, it is quite enough for learning Pāli. It is in
fact incredible to traditional students, who may undergo more
hardship without using such tools.
From this chapter on, we will see the true capacity of compu-
ter-aided learning. We will realize that data processing can
give us insight about the subject in various dimensions. In
this chapter I will introduce a tool that can show you all terms
in the collection. And here is one of indispensable tools in mod-
ern Pāli learning.
I call this tool Simple Lister. It is ‘simple’ because it just lists
terms and does something upon the list. However, because the
tool is really powerful, you have to understand some strange
options to get most of its capacity.
What is counted as ‘term’ here is simply a token. It is not a
Pāli word in strict sense. A token, in the domain of computa-
tional Pāli, is a chunk of character string without in-between
separators. So, a portion with punctuation marks will be split
with those marks resulting in standalone tokens. That ex-
plains why nti is counted as a term here. All terms or tokens
were prepared and put into the program’s database. All of
them are extracted, or technically called ‘tokenized,’ from the
72
CSCD with their number of occurrences (frequency). All to-
kens are also normalized into lowercase letters.
Simple Lister is one of the tabs in the main window. It
can be opened as a separate window by Bars button, or by the
Collection menu. At first, it will show the top-most frequent
terms. We can see the summary by using Σ button. The re-
sult is shown in Figure 13.1. With Check-double button, you can choose
between term and document summary. In the picture, term
summary is shown.
Figure 13.1.: Simple Lister window with term summary
In the result table, each row shows the term, its frequency,
and length. The length is Pāli-sensitive, meaning th, for ex-
ample, has 1 character long. So, you see tattha in the list has
length = 5.
By default, the result is retrieved only 500 rows, ordered by
top frequency. This makes the starting time is very fast. You
can set the number of maximum rows by a drop-down option
in the tool bar, ranging from 50 to 1,000,000 rows. The less
you select the faster you get the result.
The table can be sorted by a specific column. You just click
on the column header you want. But remember that the sort-
ing is done only on the existing data limited by the maximum
73
13. Simple Lister
rows chosen. It does not retrieve new data.

At the bottom, there are text group selectors (V = Vinaya,
S = Suttanta, A = Abhidhamma). You can include text groups
so that you see only terms in these groups. Result from select-
ing all groups is fast, as well as an individual group. A combi-
nation of groups, not all of them, may take a little more time
to process. When all groups are selected, and the maximum-
rows is set to 1000000, the result is the list of all terms in the
collection.
Another useful option is frequency range, the drop-down be-
fore the maximum-rows option. By default it is set to >= 1
(greater or equal to 1), meaning all in the range. You can
select other criteria to see only what you are interested in.
For example, by selecting = 1 (equal to 1) you will see only
terms that appear only once. These are more than 500,000
terms in the whole collection. That explains why 550000 is in
the maximum-rows option, not shear 500000. Playing around
with this option can bring you some insight about statistics of
the textual data.
Now you can apply all the options mentioned to find the
longest term in the collection. First, select all text groups.
Second, select 1000000 maximum rows. When the result is
shown, click the header of column Length twice. Figure 13.2
shows the result of what I have done myself.
If you choose the full frequency range (>= 1) like me, you have
to wait a little long.1 It will be faster if you choose = 1 instead
(very long terms are supposed to appear once), because less
data will be processed.
Simple Lister also incorporates a powerful search tool. The
search result comes up immediately after you enter the first
character, or you can drag a word from other places and drop it
here. You have four search modes: (1) Simple filter, (2) Using
? and *, (3) Regular expression, and (4) Filter by meter. These
options can be selected by Check-double button behind the search input.
In the first three modes, the searching is done directly in the
database. So, you do not worry that whether your selected
maximum-rows is sufficient. In contrast, the meter search is
done upon the retrieved data. You have to choose a suitable
1 It is processed under 10 seconds each step in my computer, pretty fast in
fact. It is long because the result does not come up right away.
74
Figure 13.2.: Top longest terms in Simple Lister
number of maximum rows.

The input acceptable in filter-by-meter mode is (case-sensi-
tive): 1, 2, 4, l, g, n, s, j, b, m, N, S, J, Y, B, R, T,
M, and L (see Chapter 9 for more information).2 Be care-
ful with 2 here. It means either g or ll. Figure 13.3 shows
the result of gggl with 500 terms retrieved, comparing to Fig-
ure 13.4, which uses the input 2221 instead. So, a suggestion
when you use the meter search is always use g (not 2) when
you search with raw meter (1 and l bring the same result).
The last feature I will talk about is grouping. Grouping is an
action upon the retrieved data. It groups terms by similarity.
For example, you can see that how many terms starting with
each letter by setting the start number to 1 (see Figure 13.5).
There are some limitations to be aware of. First, characters
here are not Pāli-sensitive. This means under the group of
‘p-’, for example, terms starting with ‘ph-’ are included. And
Second, the operation is done upon the existing data, so you
have to select a proper number of maximum rows.
Similarly, you can see that how many terms ending with the
2 The weight 6 and 8 are excluded from the process because they are difficult
to implement. You have to be specific. For example, using 24 or 42 for 6 and

use 44 for 8.
75
13. Simple Lister
Figure 13.3.: Filter by meter ‘gggl’ in Simple Lister
Figure 13.4.: Filter by meter ‘2221’ in Simple Lister
same letters by setting the end number. Figure 13.6 shows

the result of two-letter endings. Remember that the hyphen
(-) includes zero or any number of characters. So, ‘-ti’ means
‘ti’ itself and everything ending with ‘ti’. You can certainly
set both the start and end number to see a particular effect.
Please experiment.
76
Figure 13.5.: Grouping by the first letter in Simple Lister
Figure 13.6.: Grouping by the last 2 letters in Simple Lister
77
Part IV.
Advanced Search Tools
78
14
Lucene Finder
In Chapter 11, Document Finder, a tool that can locate re-

quired documents, is described. It is simple and handy, and
it works fine in common situations. If the tool is not powerful
enough so that you still cannot find what you want. Lucene
Finder is the final answer to all sophisticated search. It is re-
ally easy to use and powerful, but you have to learn some new
expressions of the search query.
Figure 14.1.: Apache Lucene’s logo
This tool utilizes Apache Lucene 9 as the search engine. You

cannot use it yet in the first run. You have to build a Lucene in-
dex first. Lucene index is a group of files generated by Lucene
with options specified by the users. The files reside in a direc-
tory (data/index/main by default). The users can build indices
with a variety of options as many as they need, but only one
index is selected to use at a time.
79
14. Lucene Finder
14.1. Options for indexing
Lucene Finder can be opened by its button in the tool bar, or by

the Collection menu. At the title bar, the name of index’s di-
rectory is shown, together with number of documents indexed.
So, at the first open, you should see [main:0]. It is better to
learn the options before you build the index (see Figure 14.2).
Figure 14.2.: Options for building Lucene index
(1) Text group The first option you should consider is which
group of text you will work with. There are several options
available, running from a single set of text to the whole col-
lection. In general, indexing the whole collection is suitable
for most use cases. If you have a particular purpose to work
with a set of text, you can make the index for it and save in a
different directory.
(2) Keep capitalized terms This option allows capitalized terms

to be indexed and searched. Normally, we lowercase all tokens
when indexing. That can save space and reduce search com-
plexity. With English corpora, it is a common practice. But in
Pāli, particularly in the collection we have, capitalized terms
are those starting sentences. So, they are embedded with ad-
ditional meaning. Keeping capitalized terms, therefore, al-
lows us to do a more refined search. By default, this option is
not set.
80
14.1. Options for indexing
(3) Include numbers In the body of text, there is no use of num-

bers in Arabic form, except paragraph numbers. Numbers can
also be found in editorial notes. When this option is set, you
can search numbers in the text. This has little use in most
contexts, so this option is turned off by default.
(4) Include field ‘bold’If you have ever opened a document in

the program’s viewer, you will see that most of them has por-
tions of text running in boldface. I call this field ‘bold’ (learn
more about fields in the next section). The bold text signals
that there are explanations on that part in the text’s commen-
taries/subcommentaries. Should we include this? A typical
answer is ‘No.’ Because the bold part of text is already col-
lected in other fields. You lose nothing when ‘bold’ is excluded.
However, when you really need to search text in ‘bold,’ this
field has to be included in your index.
(5) Length exclusionYou can exclude terms by their length. We

have three options here: == 1, <= 2, and <= 3. The first option
is the default, meaning all single letters are excluded. If you
choose the second, all terms with 2 letters and 1 letter long
will be excluded. So, you cannot find ‘ti’ or ‘ca’ or ‘vā’ in this
case. Cautiously use this option, or else you will lose a good
number of words.
(6) Use stopwords Finally, you can exclude specific terms by

adding them into a list. Technically, we call such terms stop-
words. The stopword file is hardcodedly set to data/rules/stop-
words.text. You can edit it by the menu provided, or using an
external editor. By default, this option is turned off.
Once you finish your option choosing, then you hit the Build
button to create the index. The program will ask you for the
output directory. You can create a new one at this step. After
that the program will do its job, and you need to wait. Index-
ing is heavily resources-consuming, so you should not do other
things meanwhile. It will not take long.1 For the whole collec-
1 In my old dual core 32-bit laptop, it takes only 2 minutes for the whole
collection with default settings.
81
14. Lucene Finder
tion, if you see 2698 (the number of all documents) in the title
bar, the index is successfully built.
14.2. Description of fields

Thanks to XML format of files in the corpus, we can search
by selecting relevant fields. When the texts are indexed, not
the whole bunch of them is processed, texts are pigeonholed
into fields. Fields were pre-organized by the collector of CSCD.
Table 14.1 show all fields used in Lucene Finder. Selection of
fields can be done by Tasks button in the second tool bar.
Table 14.1.: Fields used in Lucene Finder

Field Group Description
bodytext body the main body of text
center text with center alignment
indent text with an indent, outside bodytext
unindented text without indent, outside bodytext
nikaya headings e.g. Vinayapiṭake
book e.g. Pārājikapāḷi
chapter e.g. Pārājikakaṇḍaṃ
title e.g. Paṭhamapārājikaṃ
subhead e.g. Vinītavatthu
subsubhead rare, arbitrary, e.g. Nidānagāthā
gatha1 gatha the first line of a stanza
gatha2 the second line of a stanza
gatha3 the third line of a stanza
gathalast the last line of a stanza2
note note editorial notes, text in […]
bold bold text in boldface
2 When a stanza has only 2 or 3 lines, ‘gathalast’ is always the last. The
unused lines are skipped. That is to say, ‘gatha1’ is always present, ‘gatha-
last’ is almost if the stanza has more than one line, and ‘gatha2’ and ‘gatha3’
appear only in long verses.
82
14.3. Lucene simple search
Sometimes it is hard to tell, if we do not look into the XML

files, one field from another. For example, field ‘center’ looks
similar to those in the headings group, but it normally con-
tains a longer portion of text, whereas text in headings is typ-
ically short, if not only one compound word. Knowing all these
can help you search more effectively. You can, for instance,
search text only in verse form, or more specifically in any line
of verse form.
Only ‘bold,’ if included in the index, forces exclusively one-
field search. Other fields can be selected in combination. To
ease the user, fields are grouped so that you can select multiple
fields at once. So, you see two modes of field list: simple and
detailed mode.
14.3. Lucene simple search

After you build the index and know how fields work, then you
can use the search function. Only thing you have to do is enter
some word and hit the Enter key, or Search button. I will show
you an easy case first. If you know exactly what you want to
find, and know which fields it resides, you should enter the
full word and select the fields accordingly. This leads you to
the result immediately. Figure 14.3 shows the result of finding
‘dhammacakkappavattanasuttaṃ.’
Figure 14.3.: Simple one-term search in Lucene Finder
Options concerning search result you should know are the

number of maximum results (10–100), the field selector (Tasks
button), and the text fragments (GLASSES button). It is really un-
necessary to use a high number of search results, except when
83
14. Lucene Finder
you want to see a lot of them. The main reason is the results
are ranked by scoring, and the most relevant result is shown
first. The method of scoring is the Lucene’s default and mathe-
matically complex. You do not need an explanation. Just keep
in mind, a higher score means more relevant. Sometimes, the
rank of scores does not make sense to you, or even to me, be-
cause we do not know its internal conditions. The output, how-
ever, is mostly reliable.
The text fragments are parts of the result that match the
query, showing with highlight and their context. This is a use-
ful feature, but unfortunately buggy. If you enter full words,
you are likely to see the fragments. If you use wildcards (see
below), it is less likely you will see them.3
Once you have search results, you can open the documents
shown by using context menu (right-click), as shown in the
picture.
Figure 14.4.: Simple two-term search in Lucene Finder
A search query is not limited to one word. You can enter

multiple words as many as you want. The result in this case
will be the documents that contain all the words you enter
(see also logical operators below).4 Figure 14.4 show the result
3 Lucene has its highlighter module, and we use it. But sometimes the
highlighter misses the result completely. I compensate this with my imple-

mentation. It works fine, but not perfect. That is why you still see nothing in
text fragments sometimes, even if the query is accurately matched.
4 In other words, all query words will be joined with AND operator. This is
84
14.4. Lucene query syntax
when searching ‘dukkhaṃ ariyasaccaṃ’ (you might know that

the Dhammacak has this phrase). This time you see several
documents. You have to check the fragments whether that
document is what you need. Note that these two words may
not sit together, as shown in some of the fragments. By using
Angle-Double-Up and Angle-Double-Down button, you can collapse and expand all fragments
quickly. Moreover, if you feel that the fragments are too short,
you can set ‘Show whole lines’ option by Check-double button in the second
tool bar.5
In Chapter 11 we talk about content searching to find a doc-
ument with Document Finder. Using that tool is similar to
simple search here in some respect. Once the index is built,
searching with Lucene Finder is faster and more flexible. Also,
you can see inside the documents if they are really what you
want. Moreover, if you do not have an exact word to find, in
Document Finder you are in a hardship, but in Lucene Finder
there are ways to deal with a vague query, as we shall learn
in the next section.
The best practice in finding unfamiliar words is checking
them with Simple Lister first, then you can take the words
from the Lister to search here, or in Document Finder, by drag-
and-drop. Keep in mind that entering full-word query always
bring better search result.
To use Lucene effectively, you have to know its syntax. And

this is the true power of Lucene. Some parts of the explanation
below may look difficult or too complicated for most students
of Humanities who use this program. I will try my best to
keep them easy to understand. Planting with a hoe is fine,
but today we also have to learn to drive a tractor.
the default setting. If you want OR joining, you have to use the OR operator
explicitly.
5 Showing whole lines in the fragments does not work in all cases. It may
work with wildcards and regular expression, but fails with other complex
search schemes.
85
14. Lucene Finder
14.4.1. Using wildcards

Here we meet our old friends. Lucene accepts two wildcard
characters: question mark (?) and asterisk (*). The former
stands for a single character, the latter multiple characters.
We also use these in other parts of the program.
Here are some examples: Entering ‘dhamm?’ can match
‘dhammo’ or ‘dhammā’ or ‘dhamme’ but not ‘dhammaṃ.’ En-
tering ‘dhammacak*’ can match ‘dhammacakkhu’ or ‘dham-
macakkaṃ’ or ‘dhammacak-whatsoever.’
One caveat, Lucene does not allow wildcards in the first po-
sition. You cannot find ‘?hammo’ or ‘*ammo’ here. But there
is no such limitation in Simple Lister (Chapter 13) and Tok-
enizer (Chapter 15). If you really want to search in this way,
use Simple Lister to find the exact word first and bring it here,
or use Tokenizer to make a custom index and use its search
function instead.
14.4.2. Using regular expression

Regular expression enhances the use of wildcards substan-
tially, like you replace a slingshot with a machine gun. Other
parts of the program allow using regular expression as well.
The downside of this is it takes a steep learning curve. The
syntax also looks bizarre, if not terrible, to non-technical users.
Teaching how to use regular expression is not an easy task ei-
ther. However, in Chapter 20 I have a brief treatment of this
topic.
To use regular expression in a query, just enclose it with
slashes /…/. For example, entering ‘/dhamm[oāe]/’ can search
either ‘dhammo’ or ‘dhammā’ or ‘dhamme.’
14.4.3. Using fuzzy query

As explained in the Lucene’s API document, fuzzy search uti-
lizes Damerau-Levenshtein Distance algorithm. We can use
this mode by adding a tilde (∼) to the end of a search term.6
6 Optionally you can put a number after the tilde, like ‘dhammo∼1.’ The
number specifies “the maximum number of edits allowed.” The value can be
either 0, 1 or 2 (default). I cannot give you any clearer explanation of this.
Just try it yourselves.
86
This will find terms that look similar to the query.

For example, in my search of ‘dhammo∼’ I get these in re-
turn: dhammo, dhammā, dhamma, dhamme, dhammaṃ, ad-
hamma, adhammo, dammi, dhajo, etc.
14.4.4. Using proximity

Normally when we enter multiple search terms, we do not use
quotation marks (“…”). If the proximity of the terms is taken
into account, we use quotation marks. That is to say, if you
want to search exactly ‘dukkhaṃ ariyasaccaṃ,’ enclose the
phrase with double quotes (hence “dukkhaṃ ariyasaccaṃ”).7
In addition, you can specify the distance by adding a tilde
(∼) at the end plus a number. For example, entering “pana
bhikkhū”∼2 can search ‘pana’ and ‘bhikkhū’ within 2 words
apart. This can match either ‘pana bhikkhū’ or ‘pana some-
thing bhikkhū’ or ‘pana something something bhikkhū’ (see
Figure 14.5).
Figure 14.5.: Proximity search in Lucene Finder
14.4.5. Using range

This mode can be useful when searching a range of numbers.
We normally do not use this because we hardly search for num-
7 This is equivalent to “dukkhaṃ ariyasaccaṃ”∼0.
87
14. Lucene Finder
bers in the texts, except you want to find certain paragraph

numbers by building the index with numbers included.8
To search a range, use TO operator (all caps) and enclose the
phrase with square brackets, for example, ‘[120 TO 250].’
Range is not limited to numbers. It can be used with terms
as well. But it still has little use here, because range is not
Pāli-sensitive. You have to think it in English. Figure 14.6
shows some results of a search for ‘[dhamma TO dhamme].’
Figure 14.6.: Range search in Lucene Finder
As you see in the picture, every term between ‘dhamma’ and

‘dhamme’ is found, but not ‘dhammo’ because o comes after e.
But if you search ‘[dhamma TO dhammā]’ instead you will also
see ‘dhammo.’ That is not we expect.
14.4.6. Using term boost

When you search multiple terms or phrases, you can put un-
equal weight to each term to make the heavier term has higher
degree of importance or relevance. This is called boosting. It
is can be done by adding a caret (^) symbol after the term. For
example, in searching ‘dukkhaṃ ariyasaccaṃ’ if you put more
8 Paragraph numbers in the collection are not congruous. Some are put as
a range, some are missing. Some documents do not have numbers at all. This
warns you that not every number is searchable, even if it is seemingly to be
there.
88
weight to ‘dukkhaṃ,’ say, 10 times, you can enter ‘dukkhaṃ^10

ariyasaccaṃ’ (see the result in Figure 14.7, compare it to Fig-
ure 14.4).
Figure 14.7.: Term boosting in Lucene Finder
14.4.7. Using logical operators

To refine our search, we normally use multiple-term queries.
When multiple terms are present together, they have a logical
relation to each other. This relation is denoted by operators,
which we have five of them here: AND, OR, NOT, plus (+), and
minus (-). The first three must be in uppercase. By default, or
by the absence of any operator, terms are related with logical
AND, as I show you in Section 14.3 above. If you want to use
other relations, you have to explicitly use the operators.
Plus (+) means always present, minus (-) means always ab-
sent, and no symbol means optional. These two symbols work
in a similar way as AND, OR, and NOT, but have their own
meaning and use. Normally, we use the two sets of operators
exclusively, but they can be mixed.
Here are some typical uses of these operators.
This finds documents containing

(1) dukkhaṃ NOT ariyasaccaṃ
dukkhaṃ but not ariyasaccaṃ.
89
14. Lucene Finder
This finds the

(2) “dukkhaṃ ariyasaccaṃ” AND (vijjā OR āloko)
phrase “dukkhaṃ ariyasaccaṃ” with vijjā or āloko.
(3) (dukkhaṃ OR dukkhasamu*) AND ariyasaccaṃ This finds duk-

khaṃ and ariyasaccaṃ or dukkhasamu* and ariyasaccaṃ. Us-
ing a wildcard can save some typing but also can bring irrele-
vant results.
The result of this must contain

(4) -dukkhaṃ aniccaṃ +anattā
anattā but not dukkhaṃ; aniccaṃ is optional.
This finds documents having

(5) -“dukkhaṃ ariyasaccaṃ” +vijjā
vijjā but not “dukkhaṃ ariyasaccaṃ.”
14.5. Concluding remarks

Lucene is really a powerful search tool. To use it at full ca-
pacity, you have to learn how to build index with options that
agree with your specific needs. You have to know about data,
i.e. the fields that are used to structure the documents. And
most importantly, you have to know its search syntax. I use
a lot of space for this chapter because non-technical users are
usually frightened by complicated tools, so they need a friendly
guide.
What we have learned so far is just one part of Lucene’s com-
prehensive features. Lucene has many functions that can be
used in Natural Language Processing (NLP), beside the Infor-
mation Retrieval (IR) system that we utilize. Even so, we just
touch the surface if its IR capacity. The software can do much
more than I have shown you here. However, I think what we
know here is enough for Pāli learning and research. Now you
should have ideas to play around to deepen your understand-
ing both of the tool and of the language.
90
15
Tokenizer
This is the most difficult module in the program to write. I

spent more than a month to implement this alone. In a nut-
shell, Tokenizer is a mini version of Lucene. Why we need
this if Simple Lister and Lucene Finder can do the same job?1
If you find that both modules mentioned are enough for your
uses, it is not necessary to use Tokenizer. However, there are
scenarios that this module is needed or helpful:
If
(1) You can index custom documents in the Extra with Tokenizer
you have your own collection and want to search or make its
term list, you have to use this. By putting your documents in
the Extra and add them to Tokenizer, you can get the term list
and you can search for a document you are looking for.2 Even
though the search function in Tokenizer is not so powerful as
Lucene Finder, it is enough to get the job done.
As we
(2) You can add any document in the collection to Tokenizer
have seen in Lucene Finder’s options for indexing, we cannot
1 As a matter of fact, I wrote Tokenizer before Simple Lister and Lucene
Finder in the hope that I could get rid of Lucene from the program’s library
and reduce the database side. After I finished the module, I realized that I
had made a wrong decision. I thought that after all Simple Lister is needed
for its simplicity and speed and Lucene Finder is needed for its superb search
function.
2 For plain text documents, all tokens are put into the bodytext field.
91
15. Tokenizer
select documents arbitrarily. There are preset groups of texts

to be indexed. In Tokenizer, in contrast, we can select docu-
ments freely. Then we can search or explore the term list only
in the texts we are interested in.
This is a con-
(3) You can create a custom term list in Tokenizer
sequence of the previous item. When you select only a text
group that interests you, and make the term list out of it. You
can use this list in Declension Table and Conjugation Table to
check against the terms produced, alternatively to the whole
term list.
If you
(4) Capitalized terms are analyzed statistically in Tokenizer
take capitalized terms seriously, this can give you more detail
on this matter than Simple Lister or Lucene Finder. It calcu-
lates the percentage of each term found as capitalized. This
may look trivial to other languages, but in Pāli it is informa-
tive. You can know which terms are normally used as sentence
starter. In Tokenizer, you have no option for normalizing cap-
italized terms, because it is always kept as such.3
Figure 15.1.: Tokenizer window in full
Now I will explain briefly how to use Tokenizer and what

else you should be aware of. First, the module is one of the
tabs in the main window, which is known by other parts of the
3 As a result of this implementation, I drop the calculation of terms’ length
from this module. For that information, use Simple Lister instead.
92
program as the ‘main Tokenizer.’ We can open it as separate
windows as many as we want, but only the main Tokenizer is
the target of the addition by context menus or tool bars.
This leads us to two ways of adding documents into Tok-
enizer: by context menus (or tool bars) and by drag-and-drop.
There are three places that provide you a document list: TOC
Tree, Bookmarks, and Document Finder. If you use the con-
text menu (right-click) from these to add documents, the ad-
dition will be done in the main window’s Tokenizer tab. But
you can drag-and-drop freely from these three modules to any
Tokenizer opened.
Term filtering has the same function as that in Simple Lis-
ter, so as the field selector in Lucene Finder. Search pane is
not visible by default. You have to click SEARCH button. You can
drag-and-drop a term in the list to the search text field. A
multiple-term query uses logical OR relation (unlike Lucene
Finder which uses AND by default). You cannot use Lucene
syntax here. Only terms in full form are accept as valid query.
No special symbols are used. The search result is similar to
that of Lucene Finder, with slightly different display and op-
tions.4 Figure 15.1 shows Tokenizer window in its full form.
There are many things I have not talked about. I leave them
to the users. You should explore by yourselves what else you
can do. Try right-clicking here and there, and set various op-
tions to see their effect.
4 Score calculation in Tokenizer and Lucene are different, so you should
not expect the same ranking in both. For Lucene, I do not know exactly. In
Tokenizer, I just simply use logarithmic TF-IDF (see Wikipedia for more in-
formation).
93
Part V.
Miscellaneous Tools
94
16
Pāli Text Editor
In this part we will examine general tools that can be used

for various purposes. The first I introduce here is Pāli Text
Editor. It can be opened by PENCIL-ALT button in the main tool bar. To
open the editor with an existing text file, use File-Upload button instead.
The function of these two buttons can be invoked also by the
File menu. This is a simple text editor with basic functions,
but supports to Pāli language are added, e.g. we can type in
Pāli text easily with built-in input methods (see Section 2.4).
I will not explain the basic functions of the editor, which
are familiar to most computer users and easy to learn. What
makes the editor useful to Pāli learners is text processing tools
provided. You can see these in menu Tools (see Figure 16.1).
By using tools, you can convert text from Roman script to
other scripts or vice versa. You can remove diacritic marks
from Roman text, decompose them, or re-compose them. You
can calculate raw meter, send a portion of text to prosodic an-
alyzer (see Chapter 9), or Pāli Text Reader (see Chapter 18).
You can change the text to lowercase or uppercase. You can
sort a list of terms by Pāli order. Finally, you can change Pāli
characters into TEX format or vice versa. I added the last func-
tion for my own use because I use a lot of such conversion in
my book writing.
The best way to understand all these is to play with them.
One caveat, some operations are undoable, particularly when
you operate on a selected portion of text, but some are not
95
16. Pāli Text Editor
Figure 16.1.: Text-processing tools in Pāli Text Editor
(even with selected text). So, it is advisable to save you doc-

ument before you use any tool, unless you are familiar well
with the behavior of the actions. Some critical operations do
not make change to the original text. They create new text in
a new editor instead. Script conversion is a marked example.
96
17
Batch Script Transformer
This tool helps you do script transformation for multiple files

at a time. This can save time if you have many Pāli text files
needed the conversion. The tool is simple and straightforward.
It works only with text files. You can open the transformer by
menu File>Batch Script Transformer.
First, you have to add files by hitting the button provided,
select the output script you want, then hit the Convert button.
That is all. Figure 17.1 shows the result from my playing.
Figure 17.1.: Batch Script Transformer window
There are some other things you should be aware of. You can
convert from Roman script to any of other scripts, but other
scripts can be converted only to Roman. You can set the out-
put folder by the button provided, otherwise you have to use
97
17. Batch Script Transformer
suffix added to the output files’ name. The default suffix is

‘_converted.’ You can change this by its button. You can choose
whether numbers are included in the conversion or not. This
option works only in the transformation from Roman to other
scripts. The Reset button is used after a transformation is
successfully done so that you can do it again, possibly with a
different output script.
98
18
Pāli Text Reader
To address this question, “What is the best way to make Pāli

text reading easier?,” I have developed this module. In the
document viewer, as we have seen in Chapter 12, we have an
embedded dictionary that can give us meaning of a word at
a time. This can be helpful when you scan the text and get
stuck with some words, but it is too cumbersome for deliberate
reading.
Here is how the reader works. First, you open the reader
window by BOOK-OPEN button in the main tool bar, or in the Collection
menu. Then, you select a portion of text, either in the HTML
viewer or the text editor or even an external editor, copy it and
paste into the reader. In the HTML viewer and text editor, you
can open the reader directly by their context menu. To paste
the copied text to the reader, you can use Paste button in its tool
bar, or press Ctrl-V.
Now I will explain the conceptual idea behind the module.
Please read the following carefully, especially when key terms
are introduced.
Once the reader gets the raw text, the reader splits the text
into sentences. What is counted as a sentence here is any sec-
tion that starts with a capitalized term. It can be very short
(only one word), or very long (hundreds of words). That is to
say, sentences are determined by the compilers of the text, not
options able to set by the program.
As we have immutable text in the collection, when we paste
99
18. Pāli Text Reader
portions that have the same arrangement into the reader, we

will always get the same individual sentences. By this way,
each sentence is unique by its components. Even if two sen-
tences are the same in grammatical sense, they can be differ-
ent to each other if their arrangement is different (punctuation
symbols do not count).
When sentences are strung together, they constitute a se-
quence. So, when you paste a portion of text, you work with
a sequence that have several sentences. That is to say, the
reader has no concept of paragraph or passage or discourse.
Its fundamental element is sentence; a collection of those ele-
ments is sequence.
Now I hope you understand our technical terms here. If not,
you may not fully understand how the reader, also Sentence
Manager, works. Once you have a sequence in the reader (it
must have at least one sentence), you can save it, along with its
sentences, and load it afterwards. Since sentence is the basic
unit, sequences can share the same sentences. When several
sequences are saved into a same place, if some sentences al-
ready exist, you will be asked to replace them or not. That is
because the program save only one instance of a sentence. The
sentence’s identifier is call hash.1
By the infrastructure described above, we can also add trans-
lations, and/or explanations, to a sentence. By this model,
sentences are separated from their contexts. We have the sen-
tence pool. When we read a sequence, it picks its sentences,
which may have translations embedded, from the pool. You
may see this as a drawback or limitation. However, this treat-
ment is suitable for the repetitive nature of the Pāli text, I
think.
Now you know we can add translations to a sentence. The
last key word you have to know is variant. A variant is a varia-
tion of translation identified by its author or related informa-
tion. You can see a variant simply as an author of translations.
But one author can also be represented by several variants, or
seen as versions. By the use of variant, we can add transla-
tions, including any native language2 , as many as we need to a
1 Hash is a string of hexadecimal numbers. It is calculated from its corre-
sponding sentence. Each sentence has its own hash. A duplication is possible
but very very unlikely. Technically, I use MD5 digest calculation here.
2 By native language, I mean the language or locale of the users recognized
100
single sentence. This can demolish the illusion of one correct
translation.
In the program bundle, I have already added many sen-
tences and several sequences. With these examples, you can
work further by your own. Figure 18.1 shows a sentence in the
reader with translations.
Figure 18.1.: A sentence with translations in Pāli Text Reader
I will not explain in detail about how to use particular com-

mands in the menu or tool bar. You can learn those by your-
selves. But some points are worth mentioning. When you see
the light bulb turns green, it signals that the sentence has a
translation. You can open the translation pane by Language button,
or hit Ctrl-T. The translations can be shown one variant at a
time, or all of them at once (use * button).
On the status bar, a summary is shown. This can tell us how
many sentences in this sequence. For example, 565/589 means
we have 565 unique sentences and overall 589 sentences (i.e.
some are repeated). The number of sentences having transla-
tions is also shown in the same manner, e.g. 332/345. And the
number of variants used in this sequence is shown in the last
part.
Translations can be edited and added here by PEN-FANCY button (only
when only one variant is shown at a time). If you want to add a
new variant, you have to add the variant in Sentence Manager
first (see Chapter 19), and then use menu Sentence>Update
by their system. So, fonts for that language are always available.
101
variant list. For a massive addition of translations, using

Sentence Manager is more convenient.
Having translations of the text is helpful to text reading, but
it is not the best way to learn the language. So, I always en-
courage learners to read the text directly. And here there are
tools to facilitate the reading. These include meaning lookup
for each term (using CPED), custom dictionary lookup3 , recon-
structing iti, automatic sandhi cut (according to pre-defined
rules), and recognition of declensions of pronouns, numerals,
and some irregular nouns/adjectives (see the options in Check-double but-
ton).
One option worth talking about here is the reconstruction of
iti. When iti is marked out in the text, for example, vilapi’nti,
the sandhi term will be tokenized into two words, i.e. vilapi
and nti. When the option of Reconstruct iti is turned on, we
will get vilapiṃ and iti.
Another example, passāmī’ti yields passāmī and ti. And if
the option of Shorten vowel before iti is also turned on, passāmī
will become passāmi, and ti becomes iti. That is what we ex-
pect when we learn in the class. But things do not go that
easy. In some instances, shortening the vowel does not make
the right word, and the program does not know that. So, you
have to make a decision whether you should use the option or
not. Either way is not perfect. For the terms that seem to have
iti but it is not marked out, e.g. abhinandunti, you can cut the
it (or nti in this case) out manually by editting the sentence
(see below).
You can edit the custom dictionary and the sandhi list by
the Edit menu.4 To enable detail mode, you have to press GLASSES
button, or hit Ctrl-I (use Ctrl-U to return to simple mode).
Figure 18.2 shows a sentence analyzed in detail.
As you see in the picture, ‘Yo’ is a pronoun in nominative
case, so the reader recognizes as such. The reader also recog-
nizes ‘rājāno’ as an irregular noun. Other terms marked with
‘[CPED]’ has a definition in the concise dictionary. For those
3 The custom dictionary has more priority than CPED. We can add new
words or replace the old ones with it. However, there is no option to turn the
custom dictionary off. You have to comment out the entries when you edit the
file.
4 The two files are hardcodedly located at data/rules/dict.txt and
data/rules/sandhi.txt respectively.
102
Figure 18.2.: Detail mode in Pāli Text Reader
with ‘[CPED*],’ the exact definition is not found, so the nearest

is shown instead. As you see in theyyasaṅkhātaṃ, the mean-
ing is not quite relevant. We can make it better by editing the
sentence (use PEN-FANCY button in the tool bar or Ctrl-E), as shown in
Figure 18.3.
Figure 18.3.: A use of edit in Pāli Text Reader
In the example, I open the edit pane and insert a hyphen

(-) in between the word, so we get theyya-saṅkhātaṃ. After
hitting the Submit button, you will see the result as shown. So,
you can get a better definition by editing the sentence, either
splitting a word, adding a new one, deleting some, or inserting
some symbols (adding two hyphens produces a dash).
An explanation about edit is needed. Edit is another field
103
which is saved together with the original text. When you edit
a sentence, you just edit this mutable field, not the original.
This means what is shown in the reader is the edit, whether it
is changed or not. You can edit the sentence any way you want
and save it, the original version is intact, so you can restore the
original whenever you need. But remember that, each sen-
tence has only one instance when saved. So, each sentence
has only one edit. When you edit a sentence that has been
edited before, the old edit is gone, the new one is saved. If
you work on edit extensively, it is advisable to save the sen-
tences/sequence in different directories. You can do this by
menu Sentence>Save this sequence as. Once you make an edi-
tion, you have to save it first, otherwise the menu is disabled.
There are many things concerning Pāli grammar that can-
not be implemented here, for example, verb form recognition.
We mostly rely on CPED in this matter. It has a good coverage,
I think. This can pave the way for future research on compu-
tational Pāli.5 As we have gone so far, it is already amazing.
I think that is enough to know to make a smooth start for
using this tool. It is really helpful, despite its complexity. You
have to learn by playing with it: make your own edits and
translations. Remember that you can ruin the given data eas-
ily. But this is not to fear. The original data is always avail-
able, either in the software’s package or in the website.
5 I make no promise whatsoever here. I have no academic interest or moti-
vation to do a comprehensive study on the subject. However, If I come up with

a way to improve the language learning, I will add it to the program. Pāli
machine translator is an impractical goal, but the best Pāli learning tool is a
viable target.
104
Sentence Manager
19
The creation of this tool is a consequence of Pāli Text Reader.
When operations upon sentences grow complex, a suitable tool
is needed. Before you read this chapter, you must understand
how the reader works (see Chapter 18), and know what I mean
by sentence, sequence, hash, edit, and variant.
Sentence Manager can be opened by Briefcase button or by the
Collection menu. It can also be opened from the reader. The
manager is the most complex tool of all, in terms of its func-
tions and components. I cannot tell and show you everything
here. I will explain only the basic ideas that you should know
when you explore the tool by yourselves. Figure 19.1 shows
the manager on the first open.1
There are three main tabs in the manager: Sentences, Trans-
lation Variants, and Merger. The first two tabs work for one
directory at a time. This means there must by one working
directory, the default is ‘main.’ You can change the working
directory by FOLDER button. The directory contains many sentence
files, several sequence files, and one variant info file.2
When the manager opens a directory, it reads all sentences
and lists them in the table, and show the number of trans-
1 Because there are many sentences to load in the first run, starting Sen-
tence Manager takes time initially. If you open the reader first, the slowness
will be of the reader instead.
2 In the implementation, I use JSON format for sentence and info files.
Sequence files are just plain text with .seq extension.
105
19. Sentence Manager
Figure 19.1.: Sentences tab in Sentence Manager
lations of each sentence, if any. When some sentences are

changed outside the manager, edited by the reader, for exam-
ple, UPLOAD button has to be pressed to update the sentence list.
You can save the whole directory into a zip file by hitting ARCHIVE
button.
On the status bar, the directory’s summary is shown. This
tell us the directory’s name, the sequence’s name (if selected),
the number of sentences, translations, variants, and sequences.
The number in the parentheses is the number of sentences
that have translations. When a sequence is selected, the num-
ber of the selected sentence is also shown.
You can search for sentences by three options: in (origi-
nal) text, in edit, and in translations (see Check-double button). You can
choose to see all sentences or the sentences ordered by a se-
quence by using LIST-OL button.
Once a sequence is selected, you can save it as a new se-
quence by Download button. You can open the sequence in the reader
by using BOOK-READER. To open only one sentence, use BOOK-OPEN instead.
When you mess up the table order by clicking its header,
to restore the order you can use the button provided. You can
delete the selected sentence from the directory by using Trash but-
ton. (Be careful, if you play with the main directory bundled.)
In the lower part of the Sentences tab, there are another
106
three tabs. In these, you can edit the sentence’s text, add/edit
translations, and see the sentence-sequence statistical rela-
tion. I will not go into the details of these functions.
Figure 19.2.: Translation Variants tab in Sentence Manager
The Variants tab is far more simpler (see Figure 19.2). All
variants and their information are shown here. You can add a
new variant (+ button) and edit its details. You can also hide
variants you do not want to see in the translations with EYE-SLASH
button. This can make the display less clustered.
If you want to delete all translations under a variant name,
you can delete that variant by Trash button. This causes all sen-
tences containing the variant to get updated. Be careful, you
cannot undo this action and a lot of data can be deleted. Make
sure you have a backup, and you are sober enough at the mo-
ment. If you just want to rename the variant, use Tag button
instead. I leave other minor things in this tab to the users to
find out by themselves.
The last tab of the manager (Figure 19.3) is really useful
practically. It can merge two directories into one with simple
steps.
First, you have to create more than one sentence directories.
You can test this by selected a sequence and save it in a new
directory. Do it again with another sequence. Then you load
these two directories into the Merger, left and right. Now you
107
19. Sentence Manager
Figure 19.3.: Merger tab in Sentence Manager
have to think whether you should merge variant information

or not. And think further, if a crash occurs (two directories
contain the same sentence), which one you want to keep its
data. The options to set lie in the lower part. Once you finish
the settings, hit the Merge button. It will ask you the output
directory. You should create a new one or select an empty one.
When done, you will see the magic.
To understand all the things we talk about here, you have
to roll the sleeves up and get your hand dirty. I have prepared
enough data you can play with. Have fun.
108
Quick guide to regular expression
20
It is probable that most users of this program know nothing
about regular expression. If you are not in a computer science
department, or leaning about formal language theory, it is un-
likely that you come across the term. But if you use computer
a lot, particular when you use programs that have advanced
search functions, the chance is good that you meet it.
What is it, then? Regular expression is a technical term,
so it will be distracting if you ask for its meaning. To put it
down to earth, regular expression is an enhancement of wild-
card pattern that can make string matching more effectively.1
Learning about regular expression is not easy. Many books
about it have several hundred pages. So, the topic is really too
big to discuss here.
However, I think we do not need to know all of its functions.
That is the reason I add this chapter to introduce the users
this powerful technique. Once you realize its capacity, you
may need to learn it more. So, in this chapter I will show
you some uses of regular expression in searching. I select only
easy techniques that can be applied to our Pāli search (see Ta-
ble 20.1). Many things not included here may or may not be
used in the program.2 You have to test them yourselves.
1 This is my definition to make it relevant here. Applications of regular ex-
pression are vast in computer science, and lesser in linguistics. To dig further,
see https://en.wikipedia.org/wiki/Regular_expression.
2 Not every technique can be used here because regular expression itself
109
20. Quick guide to regular expression
To understand regular expression (regex, from now on), you

have to know how to use wildcards first. Basically, we use
question mark (?) and asterisk (*) for this function. Tech-
nically, we call these meta-characters—characters that repre-
sent other characters. Here, ? means any one character, and
* means any characters including none. (These two meanings
are not used in regex.) For example, p??? can match pana or
puna or pitā or whatever that starts with p and have totally
four characters long. Whereas pi* can match pitā or pitaro or
anything starts with pi including pi itself. We can see that
even a simple use of wildcards can ease you search signifi-
cantly. But regex has much more than this to offer. First, we
have to forget about ? and * here because in regex they have
different meaning.
Table 20.1.: Some uses of regular expression

Pattern Meaning Example Result
. (dot) any character p... pana, puna,
pitā, and so on3
\d4 a digit \d\d 00, 01, …99
\D a non-digit \D anything but a
digit
\s a whitespace kho\spana kho pana
\b a word boundary \biti\b iti as a whole
word5
\B a non-word \Bti\b anything ends
boundary with ti
has several implementations. Even in our program, four parser engines are
used: Java (in Tokenizer, the Editor), JavaScript (in the HTML Viewer), H2
database (in Simple Lister), and Lucene (in Lucene Finder). So, a pattern
used in one place may not work in others, or may need an adjustment.
3 A dot in regex is equivalent to ? in the wildcards.
4 In the Text Editor, you can use this verbatim. But in the HTML Viewer,
you have to double the backslash, so use \\d instead. This is true to all pat-
terns that use backslashes.
5 The word boundary is not Pāli sensitive. So, a non-English character is
also counted as word boundary. For example, you may also find āiti in this
case, if there is such a word.
110
Table 20.1: Some uses of regular expression (contd…)
Pattern Meaning Example Result
\S a non-whitespace eta\Savoca6 etadavoca
[…] any in the class [Tt]ena Tena or tena
dinn[oā] dinno or dinnā
[^…] not in the class dinn[^oā] dinne
? once or not at all i?ti it or iti
* zero or more times i*ti ti or iti or even
iiti, etc.
+ one or more times manas+a manasa or
manassa
{n} exactly n times manas{2}a manassa
{n,} at least n times manas{1,}a manasa or
manassa
or even
manasssa
{n,m} at least n but manas{1,2}a manasa or
not more than m manassa
times
(…) grouping man(as)?o mano or
manaso
(…|…) any in grouping manas(o|sa) manaso or
manassa
^ at the beginning ^attha.* any word start-
ing with attha7
$ at the end .*attha$ any word end-
ing with attha
What you have seen in the table is just a small part of regex
that I think it can be applied to out search. Many other things
seem difficult to use with Pāli, at least in an easy way. For
those who know regex well, you may try group capturing and
back references, for example, using ‘(puna)p\1ṃ’ to find ‘punap-
punaṃ.’ I find little use of this, but it can be useful in some
situations.
6 You may use eta.avoca to yield the same result, but the meaning is dif-
ferent. Using \S stresses that there should be no space in between.

7 The symbol ^ and $ should be used only in term searching, like in Simple
Lister, because when searching in full text, the meaning of ‘start’ and ‘end’ is
context-dependent.
111
About the author
J. R. Bhaddacak holds a PhD in Religious Studies and has

professional background of computer science and engineering.
Nowadays he is an independent researcher, working alone out-
side any academic milieu. His main field of study is on re-
ligion, particularly Theravāda Buddhism as a cultural prod-
uct. Recently he has started investigating into Pāli language
with three goals in mind: first, to make Pāli more accessible
by making it easier to learn; second, to make Pāli studies more
critical by also taking modern literary theory and its kin into
account; and third, to research into computational Pāli and
produce effective Pāli learning tools. He is also the maker of
Pāli Platform, a comprehensive program for Pāli studies. By
the days of writing this manual, he lives as a mendicant some-
where in a rural area of Thailand.
113
Colophon
This document is produced by LATEX typesetting system using

TEX Live 2022/Debian on GNU/Linux. Devuan Daedalus/Ceres
(testing branch) is used to date. Main fonts used are TEX Gyre
Schola (Serif), TEX Gyre Heros (Sans) and DejaVu Sans Mono.
The fontawesome5 package facilitates the use of graphic icons.
To make the final PDF unicode-searchable, LuaLATEX is used
as the engine. Neovim is the main editor. The working ma-
chine is 32-bit Dell Inspiron N4030 (2011).
114

PP 2 Man

Uploaded by

Copyright:

Available Formats

PP 2 Man

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

PP 2 Man

Uploaded by

Copyright:

Available Formats

Pāli

This work is licensed under a Creative Commons

Pāli Platform 1 has been released since January 2020. It might

outside the sphere of computational technology may skip com-

1. Begin at the beginning 3

2. Basic operations and settings 16

II. Grammatical Tools 23

III. Pāli Collection 59

10. Browsing and bookmarking 60

11. Document Finder 64

12. Document viewer 67

13. Simple Lister 72

IV. Advanced Search Tools 78

14. Lucene Finder 79

16. Pāli Text Editor 95

17. Batch Script Transformer 97

18. Pāli Text Reader 99

19. Sentence Manager 105

20. Quick guide to regular expression 109

About the author 113

5.1. Expansion of case abbreviations . . . . . . . . . . 32

7.1. Tenses and moods . . . . . . . . . . . . . . . . . . 40

9.1. Syllable vs. weight summation . . . . . . . . . . 46

12.1. Transformation rules of Thai script . . . . . . . . 71

14.1. Fields used in Lucene Finder . . . . . . . . . . . 82

20.1. Some uses of regular expression . . . . . . . . . . 110

1.1. PaliPlatform2’s folder structure . . . . . . . . . . 8

2.1. The common local tool bar . . . . . . . . . . . . . 17

3.1. Menu Grammar . . . . . . . . . . . . . . . . . . . . 24

4.1. Letters window . . . . . . . . . . . . . . . . . . . 30

5.1. Declension table of a pronoun . . . . . . . . . . . 32

6.1. Result of main verbs when search ‘vim’ and se-

7.1. Aorist forms of pacati . . . . . . . . . . . . . . . . 39

8.1. Roots window . . . . . . . . . . . . . . . . . . . . 42

9.1. Prosody window . . . . . . . . . . . . . . . . . . . 43

10.1. TOC Tree window . . . . . . . . . . . . . . . . . . 61

11.1. Searching Dhammapada in Document Finder . . 65

12.1. Document viewer in its full form . . . . . . . . . 68

13.1. Simple Lister window with term summary . . . 73

14.1. Apache Lucene’s logo . . . . . . . . . . . . . . . . 79

15.1. Tokenizer window in full . . . . . . . . . . . . . . 92

16.1. Text-processing tools in Pāli Text Editor . . . . . 96

17.1. Batch Script Transformer window . . . . . . . . 97

18.1. A sentence with translations in Pāli Text Reader 101

19.1. Sentences tab in Sentence Manager . . . . . . . 106

1.1. A history of Pāli Platform

Nothing still happened yet. My life changed. In 2010, I be-

1.2. Why I take Pāli seriously?

plenty of time to do what I think it is worth spending the day

purpose of such knowledge is a kind of power, ability to control and maintain

tious manipulation of power through religion. That is what I

1.3. Features so far

Most of the text compiled in the Chaṭṭha

(2) Essential grammatical toolsNew Pāli learners will find many

by Vipassana Research Institute (VRI), tipitaka.org.

(3) Powerful search tools Searching is the primary goal of the

(4) Unique and powerful accessories The program also incorpo-

1.4. How to run the program

or add JRE by themselves), you can choose to download the