FIGURATIVE LANGUAGES – Figurative Language Classifier and

Translator using Artificial Neural Network with Rule-based technique

Ria A. Sagum Angelito M. Encinas John Harold C. Rugay
Polytechnic University of the Polytechnic University of the Polytechnic University of the
Philippines Philippines Philippines
[email protected] [email protected]
Rhegie R. Candelaria om
Polytechnic University of the Iverson C. Gruezo
Philippines Polytechnic University of the Mark Aldrin N. Santiago
[email protected] Philippines Polytechnic University of the
m [email protected] Philippines
om marksantiagopszxc@gmail
John Arcee F. Celestino .com
Polytechnic University of the Rocky E. Montallana
Philippines Polytechnic University of the Ferdinand D. Tomines
[email protected] Philippines Polytechnic University of the
m [email protected] Philippines
om [email protected]
Paul Kenneth C. Conde
Polytechnic University of the Gabriel Herdi F. Panag Jerediah C. Tomol
Philippines Polytechnic University of the Polytechnic University of the
[email protected] Philippines Philippines
gabrielherdiepanag@yahoo [email protected]
.com
Sean Caleb M. Ubias
Polytechnic University of the
Philippines
[email protected]

ABSTRACT
The use of figurative language is Figurative language is a language that
ubiquitous in natural language text and it is a contains or uses figures of speech. When
serious bottleneck in automatic text people use the term “figurative language”,
understanding. A system capable of however, they do so in a slightly narrower
interpreting figurative language would be way. In this narrower definition, figurative
extremely beneficial to a wide range of language refers to language that uses words
practical Natural Language Processing in ways that deviate from their literal
applications. A system of identifying and interpretation to achieve a more complex or
interpreting figurative language would be powerful effect. This view of figurative
helpful to NLP applications. language focuses on the use of figures of
speech that play with the meaning of words, meaning of the inputted figurative speech in
such as simile, metaphor and a Filipino word. With the use of quantitative
personification. research.

Figurative language can appear in 1. INTRODUCTION

multiple forms with the use of different Tayutay (figure of speech) is a literary
rhetorical devices and literary. Figurative form of expression used by words that are
language allows people to express abstract out of the definition of meaning to express,
thoughts. It creates tone and communicates delicate and understand the expression. It
emotional content, often through humor. The can also be used as a metaphorical form of
language that uses figure of speech is called expression that creates an image or is a
formidable form of expression that results in
Figurative language it is an effective the sole effect. In Filipino and English
way of communicating an idea that is not figurative language simile is the most
easily understood because of its abstract common and at the same time the most
nature or complexity. useful as it can be used in everyday use and
even in literary works
This study aims to enhance the past
study which is the system will identify the Language can be defined as a means
figurative language that is inputted by the of communication of human life. People
user and interpret its meaning. This system need language to communicate, to interact
can help them improve their thinking in and to get information from the other
skills in terms of figurative language (simile, people. Language is also used to express
metaphor and personification), how they are someone’s feelings or emotion and also to
differ from each other and how the express their ideas, their thoughts and their
figurative language works. imaginations, it can be spoken or written.
Written language can be found in the novel,
The system was able to improve the way newspaper, poem, and magazine. Spoken
user thinks, their skills and the ability to language can be found in the song, speech,
determine figurative language. The system and conversation. As human beings, we
was able to translate and interprets the cannot separate ourselves from involvement
of social communication and interaction, or sounds) with a meaning. Grammars are
which certainly makes ourselves impossible ways to generate signs from more basic
to live without language [1] (Ervina and signs. Signs combine a form and a meaning,
Saputri, 2007). This means people really and they are identical with neither their
rely on language as their communication. exponent nor with their meaning. Language
And it is stated that language are used to is a means to communicate, it is a semiotic
express ideas wherein such it can be found system. By that we simply mean that it is a
everything like songs, poems and everyday set of signs. In linguistics, language signs
life. are constituted of four different levels, not
just two: phonology, morphology, syntax
The use of figurative language is and semantics. [2] (Kracht, Los Angeles,
ubiquitous in natural language text and it is a CA)
serious bottleneck in automatic text
understanding. A system capable of Figurative language can tap into
interpreting figurative language would be conceptual and linguistic knowledge (as in
extremely beneficial to a wide range of the case of idioms, metaphor, and some
practical Natural Language Processing metonymies) as well as evoke pragmatic
applications. The Machine Translator is a factors in interpretation (as in indirect
system capable of identifying figurative speech acts, humor, irony, or sarcasm).
languages (particularly simile metaphor and Metonymy is by far the most studied type of
personification) and provides their literal figurative language from a computational
meanings. We create a machine translator to point of view. Computational approaches to
improve “Figurative – Literal Language metonymy are usually not limited to a
Machine Translator” which is the previous literal-nonliteral distinction. In other words,
work of Angeles, Carl Ian Jonathan B., rather than just attempting to detect whether
Evangelista, Cyrille C. and Fornis, John Ely a given linguistic expression is used
T. metonymically or not, these approaches also
try to interpret metonymical expressions in
2. RELATED STUDIES some way, as opposed to most approaches to
Languages are sets of signs. Signs metaphor (Section 3) or idioms (Section 4).
combine an exponent (a sequence of letters [3] (Rodman, et al., 2008)
 interpreting figurative language would be
Figurative language is using figures extremely beneficial to a wide range of
of speech to be more effective, persuasive practical NLP applications. The main focus
and impactful. Figures of speech such as of this thesis is on the phenomenon of
metaphors, similes, allusions go beyond the metaphor. [5] (Shutova, et al., 2011)
literal meanings of the words to give the
readers new insights. Figurative language A number of issues relating to
can appear in multiple forms with the use of metaphor, a form of expression that I believe
different literary and rhetorical devices. Its has important connections with creativity,
primary function is to force the readers to insight, and interpretation. According to
imagine what a writer wants to express. We Black, there are basically types of views
can say that figurative is a broad type of about the nature of metaphors. One is,
language. There are circumstances that in ultimately traceable to Aristotle, is that a
one sentence the figurative is indefinable metaphor involves the substitution of one
meaning it can be simile, metaphor or expression (the one used metaphorically) for
personification at the same time. The another that could have been literal. One of
language that uses figures of speech is called the most important issues that I have not
'figurative language' and "its purpose is to come to grips with is the nature of the
serve three elements of clarity, forth and mechanism for deciding whether some
beauty in the language”. However, as any attribute or set of them can be applied to
figure of speech, it has a figurative meaning; something. [6] (Ortony, 1980)
it may cause ambiguity which influences the
clarity because figurative language has some The automatic detection of figurative
specific features which make it different language, such as irony and sarcasm, is one
from nonfigurative language [4] (Clemente of the most challenging tasks of Natural
2017). Language Processing (NLP). This is because
machine learning methods can be easily
The use of figurative language is misled by the presence of words that have a
ubiquitous in natural language text and it is a strong polarity but are used ironically, which
serious bottleneck in automatic text means that the opposite polarity was
understanding. A system capable of intended. [7] (Nozza, et al, 2016)
 3. METHODOLOGY
Figurative language is using figures
of speech to be more effective, persuasive Research Design
and impactful. Figures of speech such as
Quantitative data consists of
metaphors, similes, allusions go beyond the
numbers that are obtained by using
literal meanings of the words to give the
structured and validated data-collection
readers new insights. The use of figurative
instruments and statistically
language is ubiquitous in natural language
analysed. The proponent uses
text and it is a serious bottleneck in
quantitative research to gather
automatic text understanding.
information from existing data using
sampling methods. The proponents will
A system capable of interpreting
input a word and the system will identify
figurative language would be extremely
what the type of figurative language
beneficial to a wide range of practical NLP
is. The system will use Filipino words to
applications [5] (Shutova, 2011). This study
analyze the figurative language.
aims to enhance the ability of the user or
person to identify and be more System Architecture
knowledgeable about figurative and literal
language. The system can help them
improve their thinking skills in terms of
figurative language (simile, metaphor and
personification). It can help one to be more
familiar how these 3 types of figures of
speech work, how they are used and how
they differ from each other [8] (Angeles, et
al, 2017).

Data Generation Gathering

The availability of
web, tutorials, blogs, books and other
journals that are on the internet helped the 4. PRESENTATION, ANALYSIS AND
researchers to gather information. We used INTERPRETATION OF DATA
the available list of examples from
The general objective is to create
websites, essay, poems, books and journals
a system (Machine Translator) that is
online for our sources to test our
able to improve the “Figurative – Literal
system. The researchers tested each gathered
Language Machine Translator”.
data and statistically evaluated.

1. Increase the reliability and efficiency

Sources of Data of the system.
These inputs were used for the actual 2. Increase the accuracy of identifying
testing of the proposed system. We are using the figures of speech
basic to advanced Filipino phrases as our 3. Increase the accuracy of translating
data from the internet, from the figurative languages into literal
books, magazines and any literary work that languages.
has an example of what we need.
The system produced the output
of the evaluation to present with

System Evaluation solutions to the problems raised in the

As for the study, the researchers study. The developed system “Figurative

designed a system that is used to examine Languages – Machine Translator for

how the system identifies what figure of Filipino Figurative Languages”. The

speech was used on a given statement and researchers conducted testing and

how will they figure it out. This paper serves evaluation to determine the performance

as a tool for functionality, reliability and of the system.

accuracy of the system which are the

variables of the system.
Respondent No. No. of No. of Where TP or True Positive – means
Correct Correct the number of correct sentence identified
Output for Output for and translated by the system.
classifying interpreting
FL FL FP or False Positive – means that there
1 22 4 is no figurative in a sentence but still the
2 22 4 system gives an output.
3 22 4
TOTAL 66 12 FN or False Negative – means that the

ACCURACY 88% 16% system`s output is either wrongly tagged and

incorrect which is supposedly correct
Table 1.1 shows how many correct outputs for
classifying and interpreting Filipino figurative TN or True Negative – means that there
language. is no figurative and the system said it is not
figurative language.
According to the table above (Table
Respo Precision Recall
1.1), the system has an overall performance T F T F
ndent P P N N
of 88 % in accuracy in determining which
No.
figurative language the input is and 16% in
1 3 1 19 2 75% 60%
accuracy in interpreting the input into its
2 3 1 19 2 75% 60%
literal meaning. Accuracy in
3 3 1 18 3 75% 50%
identifying the figure of speech is high TOTA 9 3 56 7 75%( 56.67%(
because the programmers used Artificial L AVG) AVG)
Neural Network to determine the weights on Table 2.1
the other side, the accuracy of translating the
given input sentence is quite low because we Table 2.1 shows the outputs of set of test

used a rule-based technique to interpret the data for which the true values are known.

meaning of the given sentence.

It shows the precision and recall for
each respondent. It has a 75% in average on
Precision while 56.67% in average on
Recall.
NO EVALUATI system and by this; they were able to
of ON OF assess the said system.
ACT ACTUAL REC PREC F-

UAL INPUTTED ALL ISION MEA

The researchers have arrived on the
REV SETS (%) (%) SURE
following conclusions. Through series of
IEW T F T F (%)
gathering data and analysis in the rules
P P N N
and how figurative works, we can say
75 9 3 5 7 56.2 75 % 64.29
that figurative is a broad type of
6 5% %
language. There are circumstances that
Table 2.2 in one sentence the figurative is
indefinable meaning it can be simile,
Table 2.2 shows the interpreting of
metaphor or personification at the same
figurative language overall results with the
time.
result of 9 TP, 56TN, 3 FP and 7 FN. The
result shows an overall percentage of
After analyzing and discussing the
56.25% for Recall and 75% for Precision.
project, the researchers have concluded
The interpretation of data resulted in that in figure of speech just a word can
a high value for True Negative; it is more make a difference in identifying and
likely to have more wrong output on determining its type of speech. The
interpreting meaning rather than giving study aims to show that learning figure
correct outputs for meaning. of speech in Filipino is as complex, if
not more complex than in the English
language. The richness of language
5. CONCLUSION
comes into play when studying such
This study “Figurative-Literal
broad language topics.
Language Machine Translator” is all
about identifying the type of figures of
speech used in a sentence and translating
it to its literal meaning. By the
experimental research method,
researchers did an experiment about the
6. REFERENCES [8] Angeles, C. et al., (2017). Figurative
Literal Language Machine
Translator.
[1] Ervina, E & Saputri, W. (2007). An
Analysis of Figurative Languages: [9] Benedek, M. (2013). Creating
“The heroes of Olympics, book metaphors: The neural basis of
three”. https://www.sc3ta.org. figurative language production.

[2] Kracht, M. Introduction to Linguistics [10] Kövecses, Z. (2002). “Metaphor A

Department of Linguistics. Los practical introduction”. Oxford:
Angeles, CA 90095–1543. Oxford University Press.

[3] Rodman, B. Narayanan, S. (2008). [11]G. Johnson, L. (1980). Conceptual

Computational Approaches to Metaphor in Everyday Language.
Figurative Language.
http://www.semanticsscholar.org.
[12]Smith, M.A., et al., (1986). Evaluation
of a simile recognition treatment
[4] Clemente, C. (2017). Literal and program for hearing-impaired
Figurative Languages. students.Journal of Speech and
www.scribd.com. Hearing Disorder, 51 (2), 134-139.

[5] Shutova, E. (2011). Computational [13] Skalicky, S. et al., (2018). Proceedings

approaches to figurative language. of the Workshop: Figurative
United Kingdom. Language Processing.

[6] Ortony, A. (1980). CENTER FOR THE [14] Kreuz, R. et al., (2007). Proceedings
STUDY OF READING: Technical of the Workshop: Computational
Report No. 154 UNDERSTANDING Approaches to Figurative Language.
METAPHORS. Massachusetts.

[15] Malgady, R. et al., (2017). Cognition

[7]Nozza, D. et al., (2016) Unsupervised and Figurative Language.
Irony Deteztion: A Probabilistic
Model with Word Embeddings. Italy.
[16] Athanasiadou, A. (2017). Irony in [23] Stowe, K. et al., (2018). Leveraging
Language Use and Communication Syntactic Constructions for
Figurative: Thought and Language. Metaphor Identification Proceedings
of the Workshop on Figurative
Language Processing.
[17] Lederer, J. (2016). Proceedings of the
Fourth Workshop: Metaphor in
NLP. [24] Wilensky, R. (1990). A critical
evaluation of commensurable
abduction models for semantic
interpretation Peter Norvig:
[18] Gordon, J. et al., (2015). Proceedings Proceedings of the 13th conference
of the Third Workshop: Metaphor on Computational linguistics.
in NLP.

[25] Honeck, R. et al., (2018). A Peacock

[19] Dunn, J. (2014). Select Multi- Looks at its Legs: Cognitive Science
dimensional abstractness in cross- and Figurative Language.
domain mappings: Proceedings of
the Second Workshop on Metaphor
in NLP.
[26] Bokus, B. et al., (2016). Psychology of
Language and Communication.
[20] Federer, R. (2016). Metaphors in
Common Speech and Tennis phrases
on NLP. [27] White, M. et al., (2017). Figurative
language in business and economics:
Metaphor and Mills.
[21] Johnson, M. et al., (2018).
Measurement of Figurative
Language: Semantic Feature Models
of Comprehension and
Appreciation.

[22] Littlemore, J. et al., (2006). Figurative

Thinking and Foreign Language
Learning.