See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/262325001

A Novel Approach to Morphological Generator for Tamil

Conference Paper · July 2010

DOI: 10.1007/978-3-642-27872-3_37

CITATIONS READS
4 691

5 authors, including:

Anand Kumar Madasamy V. Dhanalakshmi

National Institute of Technology Karnataka Pondicherry University
195 PUBLICATIONS 2,009 CITATIONS 19 PUBLICATIONS 344 CITATIONS

SEE PROFILE SEE PROFILE

Soman Kp Rajendran Sankaravelayuthan

Amrita Vishwa Vidyapeetham Amrita Vishwa Vidyapeetham
820 PUBLICATIONS 15,435 CITATIONS 237 PUBLICATIONS 821 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Rajendran Sankaravelayuthan on 23 January 2018.

The user has requested enhancement of the downloaded file.

A Novel Approach to Morphological Analysis for Tamil Language
(தமிழ் உ பனியல் ஆய்வில் ஒ திய அ கு ைற)

Anand kumar M1, Dhanalakshmi V1, Rajendran S2, Soman K P1

{m_anandkumar, v_dhanalakshmi, kp_soman} @ettimadai.amrita.edu, [email protected]
1
Amrita Vishwa Vidyapeetham, Ettimadai, Coimbatore, Tamilnadu, India .
2
Tamil University, Thanjavur, Tamilnadu, India .

Abstract

This paper presents the morphological analysis for complex agglutinative Tamil language
using machine learning approach. Morphological analysis is concerned with retrieving the
structure, syntactic rules, morphological properties and the meaning of a morphologically
complex word. The morphological structure of an agglutinative language is unique and
capturing its complexity in a machine analyzable and generatable format is a challenging job.
Generally rule based approach is used in building morphological analyzer. In rule based
approach what works in the forward direction may not work in the backward direction. The
Novel approach to morphological analyzer is based on sequence labeling and training by
kernel methods. It captures the non-linear relationships and various morphological features of
Tamil language in a better and simpler way. The efficiency of our system is compared with
the existing morphological analyzers which are available in net. Regarding the accuracy our
system significantly outperforms the existing morphological analyzer and achieves a very
competitive accuracy of 95.65% for Tamil language.

1 Introduction
Morphological analysis is the process of segmenting words into morphemes and analyzing
the word formation. It is a primary step for various types of text analysis of any language.
Morphological analyzers are used in search engines for retrieving the documents from the
keyword (Daelemans Walter et al., 2004). Morphological analyzer increases the recall of
search engines. It is also used in speech synthesizer, speech recognizer, lemmatization, noun
decompounding, spell and grammar checker and machine translation.
Tamil language is morphologically rich and agglutinative. Each root word is affixed with
several morphemes to generate word forms. Generally, Tamil language is postpositionally
inflected to the root word. Computationally each root word can take a few thousand inflected
word forms, out of which only a few hundred will exist in a typical corpus. For the purpose of
analysis of such inflectionally rich languages, the root and the morphemes of each word have
to be identified. Generally rule based approaches are used for building morphological
analyzer (Rajendran.S et al., 2001). We have implemented a novel method for the
morphological analysis of the Tamil language using machine learning approach.
2 Challenges in Morphological Analyzer for Tamil
The morphological structure of Tamil is quite complex since it inflect to person, gender,
and number markings and also combines with auxiliaries that indicate aspect, mood, causa-
tion, attitude etc in verb. Noun inflects with plural, oblique, case, postpositions and clitics
suffixes. For the purpose of analysis of such inflectionally rich languages, the root and the
morphemes of each word have to be identified. The structure of verbal complex is unique and
capturing this complexity in a machine analyzable and generatable format is a challenging
job. The formation of the verbal complex involves arrangement of the verbal units and the
interpretation of their combinatory meaning. Phonology also plays its part in the formation of
verbal complex in terms of morphophonemic or sandi rules which account for the shape
changes due to inflection. Classification of Tamil verbs based on tense inflections is evolved.
The inflection includes finite, infinite, adjectival, adverbial and conditional forms of verbs
(Rjendran.S et al., 2001). For the computational need we have made thirty two paradigms of
verb.
Compared to verb morphological analysis noun morphological analysis is less challeng-
ing. Noun can occur separately or with plural, oblique, case, postpositions and clitics suffixes.
We have classified seventeen noun paradigms to resolve the challenges in noun morphologi-
cal analysis. Based on the paradigm we classified the root words into its group. We have pre-
pared the corpus with all morphological feature information. So the machine by itself captures
all morphological rules. Finally the morphological analysis is redefined as a classification
task which is solved by using sequence labeling approaches.
3 Creating Data for supervised Learning
Nowadays machine learning approaches are directly applied to all the natural language
processing tasks machine learning approaches can be supervised or unsupervised. In super-
vised learning set of input and output examples are used for training. So, data creation plays
the key role in supervised machine learning approaches.
The first step involved in the corpora development for morphological analyzer is classify-
ing paradigms for verbs and nouns. The classification of Tamil verbs and nouns are based on
tense markers and case markers respectively. Each paradigm root word will inflect with the
same set of inflections. The second step is to collect the word list for each noun and verb pa-
radigm. Figure 1 explains the preprocessing steps involved in the development of morpholog-
ical corpus.

Figure 1: Preprocessing steps

Morphological corpus which is used for machine learning is developed by following steps:
Romanization: The set of most commonly used noun and verb forms are generated manually
for input structure and similarly the output structure is developed. These data are converted to
Romanized forms using the Unicode to Roman mapping file.
Segmentation: After Romanization each and every word in the corpora is segmented based
on the Tamil grapheme and additionally each syllable in the corresponding word is further
segmented into consonants and vowels. In segmented syllable append “–C” and “–V” to the
consonant and vowel respectively. We name it as C-V representation i.e. Consonant – Vowel
representation. Morpheme boundaries are indicated by “*” symbol in output data.
Alignment and mapping: The segmented words are aligned vertically as segments using the
gap between them. And the input segments are consequently mapped with output segments.
Sample data format is given in the fig.2 .First one represents the input data and the latter one
represents output data.”*” indicates the morpheme boundaries.

Input p-C a-V d-C i-V th th-C A-V n

Output p a d i* th th* A n*

Figure 2: Sample Data Format

Mismatching: It is the key problem in mapping between the input and output data. Mis-
matching occurs in two cases i.e., either the input units are larger or smaller than that of the
output units. This problem is solved by inserting null symbol “$” or combining two units
based on the morph-syntactic rules to the output data. And the input segments are mapped
with output segments. After mapping machine learning tool is used for training the data.
Case 1:
Input Sequence:
P-C | a-V | d-C | i-V | k | k-C | a-V | y-C | i-V | y-C |a-V | l-C | u-V | m (14 segments)
Mismatched Labels:
p | a | d | i* | k | k | a* | i | y | a | l* | u | m* (13 segments)
Corrected labels:
p | a | d | i* | k | k | a* | $ | i | y | a | l* | u | m* (14 segments)

In case 1 input sequence is having more number of segments than the output sequence. For
the Tamil verb padikkayiyalum is having 14 segments in input sequence but in output it has
only 13 segments. The second occurrence of “y” in the input sequence becomes null due to
the morphosyntactic rule. So there is no segment to map with “y”. For this reason, in training
data “y” is mapped with “$” symbol (“$” indicates null).Now the input and the output seg-
ments are equalized.
Case 2:
Input Sequence:
O | d-C | i-V | n-C | A-V | n (6 segments)
Mismatched Labels:
O | d | u* | i | n* | A | n (7 segments)
Corrected labels:
O | du* | i | n* | A | n (6 segments)

In case 2 the input sequence is having less number of segments than the output sequence.
Tamil verb OdinAn is having 6 segments in input sequence but output has 7 segments. Due to
morphosyntactic change the segment “d-C” in the input sequence is mapped to two segments
“d” &”’u*” in output sequence. For this reason, in training “d-C” is mapped with “du*”.Now
the input and the output segments are equalized and thus the problem of sequence mismatch-
ing is solved.
4 Implementation of Morphological Analyzer
Using machine learning approach the morphological analyzer for Tamil is developed. We
have developed separate engines for noun and verb. Morphological analyzer is redefined as a
classification task using the machine learning approaches. Three phases are involved in our
morphological analyzer.
 • Preprocessing
• Segmentation of morphemes
• Identifying morpheme
Fig.3 gives an outlook of the morphological analyzer system. In this machine learning
approach two training models are created for morphological analyzer. These two models are
represented as model-I and model-II. First model is trained using the sequence of input cha-
racters and their corresponding output labels. This trained model-I is used for finding the
morpheme boundaries. Second model is trained using sequence of morphemes and their
grammatical categories. This trained model-II is used for assigning grammatical classes to
each morpheme.
Input word : ப த்தான்

PREPROCESSING

Model I

SEGMENTING
MORPHEME

Model II
IDENTIFYING
Model II
MORPHEME

ப <VERB_ROOT>
Output : த்த் <PAST_TENSE>
ஆன் <3SM>

Figure 3: Schematic Representation

Preprocessing: The surface form is converted into sequence of units which is given as the
input for the morphological analyzer tool. The input word is converted into input segments
and syllables are identified. Using C-V representation consonants and vowels are
represented to the syllable.
Segmentation of Morphemes: Preprocessed words are segmented into morpheme according
to the morpheme boundary. The input sequence is given to the trained model-I. The trained
model predicts each label to the input segments.
Identifying Morpheme:The Segmented morphemes are given to the trained model-II. It pre-
dicts grammatical categories to the segmented morphemes. We have trained the system to
give multiple outputs to handle the compound words.
5 Morphological Analyzer Using Machine Learning
In morphological analysis a complex word form is transformed into root and suffixes.
Generally rule based approaches are used for morphological analysis which are based on a set
of rules and dictionary that contains root and morphemes. For example a complex word is
given as an input to the morphological analyzer and if the corresponding morphemes are
missing in the dictionary then the rule based system fails (Daelemans Walter et al., 2004). In
rule based approaches every rule is depends on the previous rule. So if one rule fails, it will
affect the entire rule that follows. In machine learning all the rules including complex spelling
rules are also handled by the classification task. Machine learning approaches don’t require
any hand coded morphological rules (Daelemans Walter et al., 2004). It needs only corpora
with linguistical information. These morphological or linguistical rules are automatically ex-
tracted from the annotated corpora. Here input is a word and output is root and inflections.
Input word is denoted as ‘W’ root and inflections are denoted by ‘R’ and ‘I’ respectively.
[W]Noun/Verb = [R] Noun/Verb + [I] Noun/Verb
5.1 Machine learning using SVM
Support vector approaches have been around since the mid 1990s, initially as a binary classi-
fication technique, with later extensions to regression and multi-class classification. Here
Morphological problem is converted into classification problem. These classifications can be
done through supervised machine learning approach.
Support Vector Machine is a new approach to supervised pattern classification which has
been successfully applied to a wide range of classification problems. SVM is based on strong
mathematical foundations and results in simple yet very powerful algorithms. SVMs are
learning systems that use a hypothesis space of linear functions in a high dimensional feature
space, trained with a learning algorithm from optimization theory that implements a learning
bias derived from statistical learning theory.
5.2 SVMTool
SVMTool is an open source generator of sequential taggers based on Support Vector Ma-
chine. Generally SVMTool is developed for POS tagging but here this tool is used in morpho-
logical analyzer for classification task. The SVMTool software package consists of three main
components, namely the model learner (SVMTlearn), the tagger (SVMTagger) and the evalu-
ator (SVMTeval). SVM models (weight vectors and biases) are learned from a training corpus
using the SVMTlearn component (Jes´us Gim´enez and Llu´ıs M`arquez, 2006). Different
models are learned for the different strategies. Finally, given a correctly annotated corpus, and
the corresponding SVMTool predicted annotation, the SVMTeval component displays tag-
ging results. SVMTeval evaluates the performance in terms of accuracy.
6 System Evaluation
Efficiency of the system is compared. The morphological analyzer system for verb and
noun are trained with 130,000 and 70,000 words respectively. This system is also tested with
40,000 verb forms and 30,000 nouns from an Amrita POS Tagged corpus (Dhanalakshmi.V et
al., 2008). The SVM based machine learning tool affords better results compare to MBT and
CRF++. Training time is very less in MBT compare to SVM and CRF++. But in testing SVM
holds good. The outputs which are incorrect are noticed and its corresponding input words are
added in the training file and trained again. This increases the efficiency of our system. This is
the main advantage of using machine learning approach to rule based approach.
7 Conclusion
This paper has described the morphological analyzer based on the new and state of the art
machine learning approaches. We have demonstrated a new methodology adopted for the
preparation of the data which was used for the machine learning approaches. We have not
used any morpheme dictionary but from the training model our system has identified the
morpheme boundaries. The accuracy obtained from the different machine learning tools
shows that SVM based machine learning tool gives better result than other machine learning
tools. A GUI to enhance the user friendliness of the morphological analyzer engine was also
developed using Java Net Beans. We are currently implementing the same methodology for
the other Dravidian languages like Malayalam, Telugu, and Kannada. Preliminary experi-
mentation gave promising results. We are confident that the proposed method is general
enough to be applied for any languages.
 References
Anandan. P, Ranjani Parthasarathy, Geetha T.V.2002. Morphological Analyzer for Tamil, ICON 2002,
RCILTS-Tamil, Anna University, India.
Daelemans Walter, G. Booij, Ch. Lehmann, and J. Mugdan (eds.)2004 , Morphology. A Handbook on
Inflection and Word Formation, Berlin and New York: Walter De Gruyter, 1893-1900
Dhanalakshmi V, Anandkumar M, Vijaya M.S, Loganathan R, Soman K.P, Rajendran S,2008, Tamil
Part-of-Speech tagger based on SVMTool, Proceedings of the COLIPS International Conference on
Asian Language Processing 2008 (IALP), Chiang Mai, Thailand. 2008: 59-64.
Jes´us Gim´enez and Llu´ıs M`arquez,2006 SVMTool:Technical manual v1.3, August 2006.
John Goldsmith. 2001. Unsupervised Learning of the Morphology of a Natural Language. Computa-
tional Linguistics, 27(2):153–198.
Rajendran, S., Arulmozi, S., Ramesh Kumar, Viswanathan, S. 2001. Computational morphology of
verbal complex. Paper read in Conference at Dravidan University, Kuppam, December 26-29, 2001.

View publication stats