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

net/publication/319090129

Morphological Generator for Kannada Nouns and Verbs

Article · April 2014

CITATION READS

1 4,457

1 author:

Bhuvaneshwari chndrashekharayya Melinamath

Sanjay Ghodawat Group of Institutions
9 PUBLICATIONS 35 CITATIONS

SEE PROFILE

All content following this page was uploaded by Bhuvaneshwari chndrashekharayya Melinamath on 13 August 2017.

The user has requested enhancement of the downloaded file.

 Volume 4, Issue 4, April 2014 ISSN: 2277 128X
International Journal of Advanced Research in
Computer Science and Software Engineering
Research Paper
Available online at: www.ijarcsse.com
Morphological Generator for Kannada Nouns and Verbs
Bhuvaneshwari C Melinamath
Department of Computer Science
University of Hyderabad, India

Abstract— The morphological generator generate morphological forms for nouns and verbs when the root word is
given. This paper describes a method of generating different word forms. Nouns in Kannada gets inflected for gender,
number and vibhakti (cases). There are 2 numbers singular and plural and 3 genders feminine, masculine and neuter.
6 cases and another set of 6 extended case like suffixes, 4 clitics. A single noun root has around 250 morphological
forms in Kannada where as a single noun root has only 2 forms in English say boy, boys. This shows the complexity
of Kannada Language. Similarly a single verb root in Kannada has around 30000 different morphological forms
where as verb root in English has 5 verb forms. Morphological generator are useful components of MT (Machine
Translation application. The input to the generator is root word followed by category and required inflection types or
all inflections. The morphological generator refers the morph dictionary to check for any morph related information
present in the dictionary. We have stored morph related information in dictionary like information regarding ‘real u’,
past participle form of verb or kinship terms, countable and uncountable features etc. These instances avoid
generator to be over general which otherwise will generate in valid word forms. Our results regarding nouns and
pronoun is more than 95% and around 85% for verbs. We have handled derivational morphology also.

Keywords— Natural language Processing (NLP), Part of Speech (POS), Expert Advisory Group on Language
Engineering Standards) EAGLES, Machine Translation ( MT).

I. INTRODUCTION
Morphology is the study of the internal structure and transformational processes of words. Words are formed from a
combination of one or more free morphemes and zero or more bound morphemes. Free morpheme are units of meaning
which can stand on their own as words. Bound morphemes are also units of meaning; however, cannot occur as words on
their own: they can only occur in combination with free morphemes. The English word jumped is comprised of two
morphemes, jump+ed. Since jump is an individual unit of meaning which cannot be broken down further into smaller
units of meaning, it is a morpheme. And, since jump can occur on its own as a word in the language, it is a free
morpheme. The unit +ed can be added to a large number of English verbs to create the past tense. Since +ed has meaning,
and since it cannot be segmented into smaller units, it is a morpheme. However, +ed can only occur as a part of another
word, not as a word on its own; therefore, it is a bound morpheme. The process by which bound morphemes are added to
free morphemes can often be described using a word formation rule.
Both analysis and generation rely on two sources of information a dictionary of valid lemmas of the language and a
set of inflections paradigms. The basic principle of morphological generation is to get forms from a root and a set of
features (lexical category and morphological properties). Generally, there are two categories of approaches to developing
a morphological generator. Approaches that use finite-state transducers (FSTs), such as Xerox Arabic analyzer (Beesley,
2003) and approaches that use rule based transformations, by ( Cavalli- Sforza, 2000)
We have developed morphological generator kannada for nouns and verbs. Nouns in Kannada gets inflected for
gender, number and vibhakti (cases). There are 2 numbers singular and plural and 3 genders feminine, masculine and
neuter. 6 cases and another set of 6 extended case like suffixes, 4 clitics. A single noun root has around 250
morphological forms in Kannada Similarly a single verb root in Kannada has around 30000 different morphological
forms. Morphological generator are useful components of MT (Machine Translation) application. The input to the
generator is root word followed by category and required inflection types or all inflections. The morphological generator
refers the morph dictionary to check for any morph related information present in the dictionary. We have developed
30000 plus words dictionary for this tag. The dictionary is tagged with hierarchical tag set. We have handled derivational
morphology also.
The remaining part of paper comprises of 5 sections. Section 2 describes the work in this area, section 3 describes
Kannada morphology, section 4 describes the proposed method, and section 5 is about performance and section 6 is
about conclusion.
II. LITERATURE SURVEY
An Lots of work is carried out on development of morphological analyzers and generators for foreign languages and
also for few Indian languages. Following is the gist of works cited in the literature. One of the most efficient approaches
to morphological analysis and generation uses finite state transducers (FST) addressed by( Mohri, 1997). There are
number of tools for the construction of FST based morphological analyzers. The best known being those developed at

© 2014, IJARCSSE All Rights Reserved Page | 1300

 Bhuvaneshwari et al., International Journal of Advanced Research in Computer Science and Software Engg. 4(4),
April - 2014, pp. 1300-1306
Xerox by( Karttunen, 1993) and by (Chanod ,1994). (Kataja and Koskenniemi, 1988) presents a system for handling
Akkadian root and pattern morphology by adding an additional lexicon component to (Koskenniemi, 1983) two-level
morphology. The first large scale implementation of Arabic morphology within the constraints of finite-state methods is
that of (Beesley, 1996). The approach of (McCarthy,1981 ) describing root-and-pattern morphology in the framework of
auto segmental phonology has given rise to a number of computational proposals.(Kay,1987) proposes a framework with
which each of the autosegmental tiers is assigned a tape in a multi-tape finite state machine, with an additional tape for
the surface form. (Kiraz, 1994) extends Kay‟s approach and implements a small work kind multitape system. (Habash,
2004). has proposed an large scale lexeme approach for Arabic morphological generation. Regarding Indian Scenario we
have attempts made by Akshar Bharati group using paradigm approach. An attempt for Kannada is made by K.N. Murthy,
in his machine added translation but the performance of the system is 60 %. From the literature survey it is clear that the
computational aspects of Kannada still need to explored. Kannada is lagging from the point of computational perspective.
Clitics aspects are need to be handled. Aspect auxiliaries and modal auxiliary kind of derivational are to be touched.
.
III. ABOUT KANNADA LANGUAGE
Kannada is a Dravidian language. This is spoken in southern India. Kannada has complex morphology. Words are
built up from roots by following fixed patterns that add prefixes, suffixes and infixes to the word. This system that
studies how words are constructed from roots, and describes the patterns they follow which in English could be called
derivational morphology. Kannada language uses 49 phonemic letters, divided into 3 groups: 13 swaragaLu (called
vowels in English), 34 vyaNjangaLu (called consonants in English) and 2 yogavaahakagaLu (neither consonants nor
vowels), anusvara, namely, “aM " and visarga, namely, “ah ". Verb formation in Kannada is complicated. For the above
verb exists ( iru) we have around 3000 verb forms. We can add any number of affixes to form complex verb root. A
single root word in Kannada can give rise to a very large number of its surface forms. The richness of Kannada lies in
the fact that significant part of grammar is handled by word morphology. Consider the Kannada verb form.

BOX 1. COMPLEX KANNADA GERUND

This Kannada word form is derived by adding 9 suffixes to the root verb tappu (wrong). The formation of this verb
form is Verb-intransitive-causative-Verbal participle-reflexive + infinitive +modal (CAP) +PRO.avaru+accusative+clitic
indefinite.
A noun boy in English has at the maximum 2 inflections boy and boys. But the same word in Kannada has around
250 forms. All these morphosyntactic aspects of the complex morphology has to adopted while generation. Hence
designing morphological generator for Kannada language is quite challenging.

Box II. Different noun Forms of Kannada noun

A. Noun Morphology
Nouns in Kannada may be distinguished for gender, number, case and clitcs.
 Gender:
Nouns referring to biologically female beings are feminine in gender, that are biologically male are masculine in
gender and nouns that are not thought to be ”rational‟(capable of thought) are referred as neuter. Sometimes young
children, small kids are treated as non-rational.

 Number:
Kannada nouns are distinguished by two numbers, singular and plural. The singular has no particular distinguishing
marker added. The plural marker ‟gaLu‟ is used for neuter nouns, ‟ru‟ is used for others. There is an exception in some
case whether gender is not specific in human noun ‟gaLu ‟ is also used as plural marker for in examples like
prajegaLu ‟citizen. Usually all masculine and feminine nouns ending in ‟a‟,‟i‟, ‟e‟, and consonant followed by enunciate

© 2014, IJARCSSE All Rights Reserved Page | 1301

 Bhuvaneshwari et al., International Journal of Advanced Research in Computer Science and Software Engg. 4(4),
April - 2014, pp. 1300-1306
u have plural marker as ‟ru‟. Some masculine and feminine are not specially marked for gender like vyaapaari
(merchants), adhikaari(officers), these words take plural marker ‟gaLu‟ which is generally used as neuter plural marker,
we are marking such words as with tag MF, to indicate dual sense. Since such word are used to represent even the
feminine gender. Some nouns have irregular plurals makkaLu ‟children‟ Some nouns have no singular forms, they have
only plural usage. Like jana ‟people‟ here janaru (people), janagaLu (people) are two plurals. aneekaru ‟many people‟.
Some noun have no usage in plural form, say gurugaLu ‟teacher‟ has no corresponding plural.

 Case system:
The case system is useful to indicate different relationships between the noun and other constituents of the sentence.
For Ex to indicate whether the noun is the ”object” of a verb (in which case it is marked for accusative case), or ”goal” of
a verb of motion (dative case), posser of something (genitive case), or the means by which something takes place
(ablative case), etc.

TABLE 1. NOUN CASES AND THEIR CHARACTERISTICS

Case/vibhakti Suffix/pratyaya Characteristics

Singular Plural
Nominative u U ru/gaLu/vu/aMdiru
(prathama)
Accusative annu nannu/Lannu/vannu/annu/yannu rannu/aMdirannu/gaLannu/yaranau
(dvitiya)
Ablative/Instrumental iMda niMda/diMda/yiMda/viniMda/LiMda/ gaLiMda/riMda/yariMda/
(tritiya)
Dative ge/ige/kke gaLige/rige/ge
nige/Lige/vige/yige
Genetive a La/na/da/vina ra/ gaLa/aMdira
(shashTi)
Locative alli nalli/yalli/Lalli/dalli/vinalli ralli/yaralli/gaLalli
(saptami)

TABLE 2. EXTENDED NOUN CASE SYSTEM

Case Suffix
Purposive Gooskara/gaagi
Comparative giMta/kkiMta
Locative_Dative oLage/allege
Similarative aMte/aMtaha/aMtha
Sociative oDane/oMdige

2.2 Verbs Morphology

Verbs are one of the most interesting and distinguishing aspects of Kannada. The verb stems in Kannada can be
divided into finite and nonfinite a finite verb ends the sentence with the exception of clitics. But the nonfinite verbs can
not stand alone and requires finite verb for ending the sentences. includes infinitives modal auxiliaries, aspect auxiliaries.
Finite verbs includes imperatives, it is in this form the Kannada verb roots are listed in the dictionary.

DERIVATIVE STEMS: These are formed by adding various kinds of derivative suffixes to the verbal or nonverbal
stem. Consider the following examples.

Derived Stem (1) tinnu „eat‟+/alu/ tinnalu, (to eat) tinnu is basic stem

COMPLEX VERBS: These are formed by adding various kinds of models to the primitive and derived stems. These
can be further subdivided into compound verbs, conjunct verbs, modal verbs and aspectual verbs. Kannada has at least
four nonfinite forms of verbs. Like the participial constructions, verbal participles, relative participles, they are
aspectually distinguished.
TABLE 3 VERB FEATURES AND CHARACTERISTICS SUFFIXES

Features Characrtistic Suffixes Example

Infinitive alu maaDu+alumaaDalu
(to do)
Imperative i,iri,oo,ee maaDu+irimaaDiri
(you do)
Negative Imperative beeDa,baaradu,kuuDadu maaDu+alu+beeDamaaDabeeDa

© 2014, IJARCSSE All Rights Reserved Page | 1302

 Bhuvaneshwari et al., International Journal of Advanced Research in Computer Science and Software Engg. 4(4),
April - 2014, pp. 1300-1306
(Don‟t do)
Optative Li maaDu+alimaaDali (let him do)
Hortative ooNa maaDu+ooNamaaDooNa (let us do)
Participle a/i/ade maaDu+uv+amaaDuva (Which will be done)
maaDa+imaaDi (after eating)
maaDu+ademaaDade ( without doing)
Verbal aspect Markers koLLu,haaku (table 4) maaDu+i+haakumaaDihaaku
(do non finite and put finite)
Causative Suffix Isu maaDu +isu maaDisu (get it done)
Conditional Suffix are, doDe maaDu+id+aremaaDidare (did they do)
Concessive Aruu maaDu +id +aruu  maaDidaruu (
even though I did)
Imprecate A maaDu +a maaDa
Emphatic clitic Ee maaDu+utt+aane+eemaaDuttaaneyee
(will he do)
Interrogative clitic Oa maaDu+uva+anu+aa maaDuvanaa
(will he do)
Indefinite clitic Oo maaDu +id +anu+oo maaDinoo
(whether he has done it)

TABLE 4: TABLE SHOWING MODAL AUXILIARIES

Modal auxiliary Meaning Example

Beeku MUST (Want) Huuvu beeku(want Flower)
kuuDadu PROH(Should not) Niivu hoogakuuDadu(you should not go)
beeDa NEG(IMP) Nanage pustaka beeDa( I donot want book)
Bahudu PERM(May) niivu barabahudu(you may come)
Aara NCAP(might not) avanu tinnalaranu( he cannot eat)
Balla CAP(capable) avanu maaDaballanu(he can do)
paDu PASS(Passive voice) haNNu tinnalapaDuttade(Fruit will be eaten)

TABLE 5: INVENTORY OF ASPECT AUXILIARIES

Aspect Marker Aspect Meaning Lexical Example
Meaning
biDu Completion Leave avanu biddubiTTanu ( he fell down)
Hoogu Completion Go anna beMditu( rice got over cooked)
aaDu Continuity Play avaru ooDaaDidaru(They ran around)
koDu Benafactive Give avanu kathe baredukoTTa( He wrote
story and gave)
nooDu Attemptive See avanu kaaphi kuDidunooDida(he drank
coffee and see
Haaku Exhaustive Put avanu dose tiMduhaakidanu
koLLu Reflexive Purchase avanu tale baachikoMDanu( he
Combed)
Aagu Finality Become avanu baMdaayitu(he finally came)
Iru Perfective Be avanu hoogiddanu(he had gone and
Stayed)

TABLE 6. KANNADA PNG MARKERS

Person Number Gender Tenses

Present Futuret Past Contigent
First Singular Masculine Eene enu Enu (y)eenu
/Feminie
Plural Masculine Eeve evu Evu (y)eevu
/Feminie
Second Singular Masculine ii(ye) e E Iiye
/Feminie
Plural Masculine Iiri Iiri Iiri Iiri
/Feminie
Third Singular Masculine Aane anu Anu (y)aanu

© 2014, IJARCSSE All Rights Reserved Page | 1303

 Bhuvaneshwari et al., International Journal of Advanced Research in Computer Science and Software Engg. 4(4),
April - 2014, pp. 1300-1306

Singular Feminie aaLe aLu aLu (y)aaLu

Plural Masculine Aare Aru aru (y)aaru
/Feminie
Singular Neuter Ade Adu itu Iitu

Plural Neuter Ave Avu avu (y)aavu

IV. PROPOSED METHOD OF DEVELOPING MORPHOLOGICAL GENERATOR

To develop a morphological generator we must consider the following factors in particular. The morphological
generator consists of following sub modules as explained in the figure 1. The process of combining morphemes involves
a number of orthographic rules that modify the form of created word so it is not a simple interleaving or concatenation of
its morphemic components.

Fig 1. Proposed Method for Morphological Generator

A. Transliteration Module: This module converts the raw text in Kannada in to roman form by using a convertor
program. If raw text is in ISCCI or Unicode then the convertor program accepts both kind of input and transliterate the
text as per our lerc notation. The input file or word should be a uninflected root.

B. RESOURCES : To develop a generator we must have some resources. These resources are
 Dictionary: A dictionary contains roots/stems, categories and exceptions, morph relevant information etc.
 Orthographic Rules Set : Morphological generators use the concatenation process while they generate a word-
form. So, we must create a set of concatenation rules for the generator according to the language. The rules
changes with word endings as shown in table 1. The alternation forms (spelling changes) of morphemes
according to the context in which they appear.

TABLE 1: SAMDHI RULES

Stem Ending
a E I o U Consonant
Real Enunciative
Glide insertion for rational n Y Y n v u dropped N
Glide for non rational v Y Y v v u dropped -

© 2014, IJARCSSE All Rights Reserved Page | 1304

 Bhuvaneshwari et al., International Journal of Advanced Research in Computer Science and Software Engg. 4(4),
April - 2014, pp. 1300-1306
C. Morphotactic Rules: These rules govern the order of suffixation. Generally the additions of suffixes to noun and verb
follow the pattern like below.

Fig 2. Morphotactic Rule for Kannada Nouns

D. Our Morphological Dictionary

The development of dictionary goes on with a specific requirement. The purpose of our development of dictionary is,
we require a dictionary of the type, lemma followed by its tag, The structure of our morphological dictionary entry is
shown in figure 2.

Word Seperator1 Tag1 Tag2 Seperator2 Morph related Information

Box 3. Structure of Dictionary Content

One entry per line. <word>||<TAG>||<TAG>:: Relevant Morph information.

Grammatical information useful for our morphology system is also stored in dictionary using the separator :: (double
colon) to help morphological generator. The word guru (teacher) is a word ending with real „u‟. But the word niiru
(water) even though ends with u but it is not real „u‟, it is enunciative. It is necessary to store such useful information for
morphological generator. We have stored such morph related information in the dictionary as shown here guru|| N-COM-
COU-M.SL-NOM-NULL: REAL u.
TABLE II. Sample Dictionary

Box 2. Sample Morphological Dictionary

D Finite State transducers

The proposed tool is developed using. Finite state transducers. Finite State Automaton (FSA) is a model of
computation consisting of a finite set of states, a start state, an input alphabet, and a transition function that maps
input symbol and current state to next state. The transducer is defined as T = (Q, L,δ, qI, F,) where Q is a finite set of
states, L a set of transition labels, qI Є Q the initial state, F ⊆ Q the set of final states, and δ : Q × L →2Q the transition
function (where 2Q represents the set of all finite sets of states). The set of transition labels is L = (Σ ∪ { ε } x (Γ ∪ { ε
}) where Σ is the alphabet of input symbols, Γ the alphabet of output symbols, and ε represents the empty symbol.

© 2014, IJARCSSE All Rights Reserved Page | 1305

 Bhuvaneshwari et al., International Journal of Advanced Research in Computer Science and Software Engg. 4(4),
April - 2014, pp. 1300-1306

Fig 3. FSM Transition for Verb Limited Version

V. EXPERIMENTS AND RESULTS

We manually generated gold standard data for nouns by choosing 2 words from a/e/i/o/u endings and 3*6*250
combinations, covering all inflections are chosen and matched that with inflected forms generated by morphological
generator. We observed that more than 95 % words are correctly matched with that of gold standard data. Similarly we
prepared gold standard data of 3000 inflections for verb. And observed that the matching performance is rate is around
85 %
VI. CONCLUSION
It is shown here that each noun in Kannada has about 250 word forms for single noun. One can observe the
complexity in morphology as compared to English, which has only three forms like say boy, boys, boy's. Similarly in
English maximum inflections for a verb 'go' are 5 in number like, go, gone, went, going, goes. But in Kannada it is
proved that more than 3000 different forms exists. One can understand the complexity of Kannada. Developing
tools like morphological analyzer or generator is quite challenging. Morphological generator for Kannada nouns and
verbs is developed here using finite state transducers. Morphological genearator is useful tool in Machine translation
applications. A dictionary of 30000 words is developed with hierarchical tag set for this task.

REFERENCES
[1] Martin Kay. 1987. Nonconcatenative finite-state mor phology. In Proceedings of the Third Conference of the
European Chapter of the Association for Computational Linguistics, pp. 2–10.
[2] L. Kataja and K. Koskenniemi. 1988. Finite state description of Semitic morphology. In COLING-88: Papers
Presented to the 12th International Conference on Computational Linguistics, volume 1, pp 313–15.
[3] Koskenniemi, Kimmo. 1983. : Two level Morphology: A general Computational Model, for Word Form
Recognition and Production. Ph.D. Thesis, Department of General Linguistics, University of Helsinki, Helsinki,
Finland.
[4] Chanod, Jean-Pierre. 1994. Finite state composition of French verbmorphology. Technical report MLTT-005,
Xerox Research Centre Europe, Meylan, France.
[5] Karttunen, Lauri. 1993. Finite state lexicon compiler. Technical Report ISTL-NLTT-1993-04-02, Xerox Palo Alto
Research Center, Palo Alto, California.
[6] K. Beesley 1996. Arabic finite-state morphological analysis and generation. In Proceedings of the 16th
International Conference on Computational Linguistics (COLING-96), volume 1, pp. 89–94,Copenhagen, Denmark.
[7] [Buckwalter, 2002] T. Buckwalter. Arabic Morphological Analyzer Version 1.0. Linguistic Data Consortium,
University of Pennsylvania, 2002. LDC Catalog No.: LDC2002L49, 2002
[8] Hopcroft, J. E., & J. D. Ullman. : Introduction to automata theory, lan guages. and computation. Reading, MA :
Addition -Wesley .
[9] Mohri, Mehryar. : Finite-state transducers in language and speech processing. Computational Linguistics
23(2):269–311. (1997)
[10] Oncina, Jose, Pedro Garc`ia, & Enrique Vidal. 1993. Learning subsequential transducers for pattern recognition
interpretation tasks. IEEE Transactions on Pattern Analysis and Machine Intelligence 15:448-458
[11] John McCarthy.1981. A prosodic theory of nonconcatenative morphology. Linguistic Inquiry, 12(3):373–418.
[12] Kiraz, G.: Multi-tape Two-level Morphology: A Case study in Semitic Non-Linear Morphology. In: Proceedings of
COLING‟94, Vol. 1 (1994) 180-186.
[13] Cavalli-Sforza, V. Soudi A., Mitamura T.: Arabic Morphology Generation Using a Concatenative Strategy.
In:Proceedings of NAACL 2000. Seattle, WA (2000).
[14] Nizar Habash. 2004. Large scale lexeme based arabic morphological generation. In Proceedings of Traitement
Automatique du Langage Naturel (TALN-04). Fez, Morocco

© 2014, IJARCSSE All Rights Reserved Page | 1306

View publication stats