Scribd
Upload
233 views7 pages

Morphological Analyzer For Amharic Language Using Rule Based and Neural Network

The document discusses verb formation and morphology in the Amharic language. It defines key concepts like morphemes, inflection, derivation, and reduplication. It also describes the different approaches that have been used to develop morphological synthesizers for Amharic, including rule-based and neural network methods. Specifically, it covers how triliteral verbs are formed in Amharic by inserting root consonants into templates and attaching subject and object markers through a multi-level process of word formation.

Uploaded by

Yared Arega
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
233 views7 pages

Morphological Analyzer For Amharic Language Using Rule Based and Neural Network

The document discusses verb formation and morphology in the Amharic language. It defines key concepts like morphemes, inflection, derivation, and reduplication. It also describes the different approaches that have been used to develop morphological synthesizers for Amharic, including rule-based and neural network methods. Specifically, it covers how triliteral verbs are formed in Amharic by inserting root consonants into templates and attaching subject and object markers through a multi-level process of word formation.

Uploaded by

Yared Arega
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 7

The scope of the thesis is limited to demonstrating the potential of a hybrid (RuleBased and

Neural Network) approach to develop an Automatic Synthesizer


morphological for Amharic triliteral2 perfective verb forms. The study excludes
beneficative and affirmative suffixes and considers only subject and object marker
suffixes
Definations
The task of morphology is thus to identify and describe the mechanisms behind this process.
Morphemes are defined as the smallest unit in language to which a meaning may be assigned or,
alternatively, as the minimal unit of grammatical analysis

Reduplication is a border case of affixation. The form of the affix is a function of the
stem to which it is attached, i.e., it copies (some portion of) the stem. Reduplication
may be complete or partial. In the latter case it may be prefixal, infixal or suffixal.
Infixal reduplications are the ones common in Amharic (Trost, 2000).
There are two productive ways to form words from morphemes: inflection and
derivation (Katamba, 1993).
Inflection morphology deals with the combination of a word with a grammatical
morpheme, usually resulting in a word of the same class as the original stem
On the other hand, derivational morphology creates new words (i.e., words with a
different part-of-speech category) by adding a bound morpheme to a stem
Mcharty stateted that there 3 layers that a verb has its elements arranged

The three universal


conventions for making associations between the prosodic templates and the
autosogementalized strings are as follows (McCarthy, 1981):
i. The meaning of a verbal lexeme is signaled at the root-tier by the
consonantal segments. Usually a verb has three consonants in its underived
lexical entry in the lexicon. Thus the root gdl in Amharic represents the
lexeme kill, which is realized by a variety of word-forms.
ii. The skeletal tier (which is also called the CV-tier) is like a potter's
template. So, it is also called the prosodic template tier. It provides a
canonical shape that is associated with a particular meaning or
grammatical function. The template CVCCVCV, for instance, carries the
22
grammatical meaning perfective. Hence gäddälä means 'he killed'.
iii. The vocalic (vowel) melody tier provides information analogous to that
carried in English by inflectional affixes like tense, aspect, number or
derivational affixes. For instance in the above Amharic perfective verb
gäddälä, the vocalic pattern ää indicates that the tense of the verb is past.
The last vowel ä indicates that the person is second person masculine and
the number is singular, as we observe in the case of English
computational morphology can be grouped into two parts:
a) Word-form synthesis and analysis;
b) Parts-of-speech (POS)-or inflectional-category determination.
Knowledge Required by a morphological synthesizer
Pullman et. al. (1988) notes that the following
are the three main types of knowledge that need to be represented for synthesizers.
i. Knowledge about the properties of the stored base forms of words:
ii. Knowledge about spelling or phonological changes upon affixation; and
iii. Knowledge about the syntactic or semantic properties of affixation (that is,
inflectional and derivational morphology).
This study will consider the possibility of generating all possible words from an input root and
suffix.
Approaches to Morphological Synthesis
1. Corpus-based approaches
a. Supervised
b. Unsupervised
Artificial neural network
Loosely modeled on the human brain, its an attempt to simulate either with
hardware or software the multiple layers of simple processing elements called neurons.
Each neuron is linked to its neighbors and learning is achieved by strengthening this
connection.
Artificial neural networks are the simple clustering of the primitive artificial
neurons. Basically, the neurons interface the real world for input and output and the
third layer is a hidden layer full of neurons.
The basic applications fall in to the following categories
i. Predication
ii. Classification
iii. Data association
iv. Data conceptualization
v. Data filtering
2. Rule-based
a. Cv-based used for stem formation the template matching handles morphological and
phonological activities of the stem components such as assimilation ,root reduction and
extension involved in stem formation.
b. Two level morphology koskenniemi’s model is two-level in the sense that a word is
represented corresponding to its surface and lexical form. After further advancements
the kimmo parser was created and has two Analytical components
i. The rules component
ii. Lexical components

Verb formation in Amharic According to different researches

BF (1978) classify Amharic verbs based on the following six criteria:


i. Consonantal skeleton (number of radicals);
Monoliteral (consisting of only one consonant);
Biliteral (consisting of only two consonants);
Triliterals (consisting of only three consonants);
Quadriliteral (consisting of only four consonants);
Quiniliteral (consisting of only five consonants).
ii. Pattern of gemination of consonants;
Type A: penultimate (second from the last) consonant that
geminates in perfect only;
Type B: are verbs whose penultimate consonant geminates
throughout the conjugation;

Conjugation means to give various inflectional endings of a verb, i.e. voice, mood, tense, number and
person.
Type C: are verbs whose penultimate consonant geminates
In perfect and imperfect only;
iii. Occurrence of vowels other than the normal ä;
iv. Identical consonants in sequence;
v. Presence of w, y, h (in the verbal root);
vi. Occurrences of initial ‘a’ or ‘t’ in base forms

Further classification of simple verbs on basis of their conjugational pattern as their phonetic
content are classified as

1. Triliterals: type A,
2. Trileterals: type B,
3. Quadrilaterals
4. Quniliterrals and sexiliterals
The minor group includes:
i. Verbs with their second and third radicals identic
ii. Verbs with

Unfinshed

Formation of perfective triradical derived stems in Amharic .. this involves prefixing and internal
changes

1. Prefixing/a-/ (transitivizer)
2. Prefix/as-/(causative)
3. Prefix/ta-/(passive marker
4. Prefix/al-/ (negative marker)

Internal changes occurs in two ways

1. The first one involves the insertion of vowel /-a-/ after the first radical of the root. It always
occurs with prefix /tä-/. It indicates reciprocity. If the subject is the cause for the reciprocal
action, the stem will be prefixed with /a-/.
2. The second one involves duplication of penultimate (the second from the last) radical
followed by vowel /ä/. It expresses that something is done again and again or
iteratively.

The above rules doesn’t apply all the time thus BF has came up with these rules

1. A verb with no antepenultimate radical cannot undergo the first


internal change.
2. A verb with initial ‘a’ does not take the /a-/ prefix.
The reduplicative stem is characterized by the repetition of the second radical in
triradicals (säbbär: säbabbär). The stems derived with prefixed morphemes are the tä-
stem, as-stem, at-stem, a-stem, taš-stem, astä-stem, tästä-stem, an-stem, tän-stem, etc.

To produce a triliteral derived stem in the perfect for type A verbs insert the
characters of the root into the template 1ä2a22ä3 or as-1ä22ä3 or tä-1ä22ä3, etc.
To produce a triliteral derived stems in the perfect for type B verbs insert the root in
to the template 1ä2a22ä3 or as-1ä22ä3 or tä-1ä22ä3, etc.
To produce a triliteral derived stems in the perfect for type C verbs insert the root in
to the template 1ä2a22ä3 or as-1a22ä3 or tä-1a22ä3, etc.

The inflectional affixes of Amharic verb stems indicate features of person, number
and gender. Amharic verbs exhibit two major types of inflectional affixes with their
verb stems;
a. Subject markers
The subject feature indicator affixes are obligatory. All Amharic
perfective verb forms cannot function without subject marker suffixes.

The following table presents subject marker suffixes in the perfective stem

b. Object markers All object indicators of Amharic verbs are suffixes.


The following table presents the object marker suffixes with their features
Summary of the possible obligatory and optional suffixes of Amharic perfective verb

the order of the


suffixes to Amharic verbs can be expressed as:
X-S1-S2; Where:
X= A perfective Stem
S1 = Subject marker suffix that occur with the perfective stem
S2= Object marker suffix that occur with the perfective stem
Mullen 1986 states there are 3 levels of word formation in Amharic language

1. Creation of stems from roots. By inserting root consonants into a template


2. Produce fully inflected words from stem or other existing words by affixation of subject
markers
3. Cliticization affixation of inflectional affixes such as object markers to the level 2 words and
compounding.
Level 2 and 3 indicate concatenative nature of Amharic

During concatenation phonological changes may occur

Vowel changes

Palatalization

Consonant changes

The following are the tables used in that database of an amhric verb synthesiszer

The tables are RootTable, TemplateTable, SuffixTable, VowelChangeTable,


ConsonantChangeTable, PalatalizationTable, and TranslationTable. This chapter will
discuss the design of each of the above tables respectively.

You might also like