Root Word Identification In Malayalam Language English Language Essay

Published: Last Edited:

This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.

Words are tools of life which is omnipresent in every language. All words in a language are unique having

their own function and meaning. The syntactic and semantic knowledge about individual words can be

encapsulated in a highly structured repository known as computational lexicon which is very essential for

Machine Translation. For designing a computational lexicon, the first and foremost task is to identify the

head words or root words in the language. The Root Word Identifier proposed in this work is a rule based

approach which automatically removes the inflected part and derive the root words using morphophonemic

rules. The system is tested with 2400 words from a Malayalam corpus to generate the linguistic information

such as the root form, their inflected forms and grammatical category. The performance is evaluated using

the statistical measures like Precision, Recall and F-measure. The values obtained for these measures are

more than 90%.


Corpus, Computational lexicon, Morphophonemic rules, Root Word, Root word Identifier


A computational lexicon plays an important role in Machine Translation since it is the place

where all the information about the vocabulary of a language is recorded for proper usage. A

Machine Translation system save enormous amount of human power and time for the translation

of one language text into another when the source and the target languages have computational

lexicons of their own. The proposed system developed a software system for automatic

identification of root words with their grammatical properties. These linguistic information about

the words are systematically stored as lexical entries in a computational lexicon.

The need for identifying the root form of a word is very important in Natural Language

Processing. A root word can be taken as a key term for searching, indexing, translating etc.

Statistical tools like frequency counter, concordance, keyword-in-context, n-gram etc. need root

form of a word to know more about the vocabulary. For example, using a frequency counter, most

frequently used word in the vocabulary can be found out. Also such information can help to

predict the correct spelling of the word, if assuming that the most frequent use of a word is

correct. Morphological analysers and generators also require root words. Lexicographers assign

root word as the head word/lexical entry in the lexicon.

Malayalam is the official language of the state of Kerala situated in the southern half of west

coast of India. Malayalam language is one among the 22 official languages of India and one

among the four major languages of the Dravidian family [1]. The influence of other languages

like Sanskrit, Tamil, Telugu, Tulu, Toda, Kota, Kodagu and Badaga is seen in phonemic,

International Journal of Computer Science & Information Technology (IJCSIT) Vol 4, No 3, June 2012


morphemic and grammatical levels of language. Malayalam is a morphologically rich and

agglutinative language. There is no distinction on upper and lower case characters.

Most of the words in Malayalam are occuring in its inflected form. For obtaining the root form of

the words, the suffixes agglutinated with them are to be removed. Also the morphophonemic

change (sandhi) occurring when a root word concatenates with a suffix should be analysed and

generalized. Sandhi rules are phonological alternations that are triggered at junctures, at junctions

of words or morphemes. Malayalam grammar has categorised sandhi rules into different types.

According to its consonant-vowel pair based categorisation, there are svara sanDhi

(svaraM~+svaraM~), svara vyanjgana sanDhi (svaraM~+vyanjanam), vyanjgana svara sanDhi

(vyanjanam +svaraM~), vyanjgana sanDhi (vyanjganaM~ + vyanjganaM~). Here svaram is

vowel and vyanjganaM~ is the consonant [1]. The morphophonemic changes occurring in the end

phonemes of the word and the initial phoneme of the suffix is used to derive morphophonemic


Morphological analysers and parts-of-speech taggers developed for Malayalam attempted to find

out the root form of the word. In our approach, no lexicon or dictionary is used. The surface

structure of the words is studied using a corpus to derive the morphophonemic rule. The Root

Word Identifier system uses these generalised rules to automatically identify the root words, the

grammatical category and their inflected forms. A corpus is created for Malayalam language

using documents from world wide web. 2400 words in this corpus are used to test the system. The

results obtained are stored in a computational lexicon having the linguistic information about

these words such as whether the word is a noun, pronoun, verb or postposition. It also gives the

inflected forms of root word present in the corpus. Performance of the system is evaluated and

obtained a high Precision, Recall and F-measure. Incremental development of computational

lexicon can be attained by using the proposed system with richer corpus as the resource.

The paper is organized as follows. The related work done in this area is presented in section 2.

Section 3 discusses about the methodology of Root Word Identifier, resource for this work and

the linguistics analysis of Malayalam words. In the same section, morphophonemic rule is

discussed. Section 4 gives the results and discussions about the work. The last section gives the

conclusion about the work.


The morphological analysis deals with the study of internal structure of words of a language

based on its grammatical category. It is the process of segmenting a morphologically inflected

word into its root word and its associated morphological components along with the features

specifying the morphological structure [2]. Even though a full-fledged morphological analyser is

not for Malayalam language, there are many attempts in this area, as discussed below.

Morphological analysis for Malayalam verbs using a hybrid approach (paradigm and suffix

stripping method) is an attempt made to attain morphological generalisation of verbs [3]. There

will be dictionary of lexical items of Malayalam, which contains lexical items, grammatical

category and paradigm type. The program compares each inflected form. The verbs are

categorized into 28 classes or paradigms based on the past tense marker. They identified around

1100 inflections of verb. Using the same hybrid approach, a Malayalam morphological analyser

using Apertium Lttoolbox is developed at Language Technology Centre, Centre for Development

of Advanced Computing (C-DAC), Thiruvanathapuram [4] as part of Machine Translation task.

Lttoolbox is available with the Apertium toolkit, which is an open source shallow-transfer

International Journal of Computer Science & Information Technology (IJCSIT) Vol 4, No 3, June 2012


Machine Translation system originated with in the project "Open-Source Machine Translation for

the Language of Spain". Lttoolbox can be customised to any language by including the required

lexical dictionary [5]. It uses the FST approach for doing lexical processing. Certain other

attempts using stochastic taggers like HMM are also in progress but they cannot give a high

accuracy for Malayalam because the language is inflectionally rich and is relatively free-word

order like Tamil.

Apart from these attempts, related works to attain the aim for developing full-fledged Machine

Translation systems are also going on in this language engineering field. Developing Named

Entity Recognizer system, Noun Phrase Chunkers, Computational Lexicon etc. for Malayalam

language are on progress.


Malayalam words can occur in its root form, inflected form, derived form, compound form and in

reduplicated form. Inflected words are formed by the affixation of grammatical features such a

case, number, tense, aspect, mood etc. to the root word. The process of separating the affixes

from an inflected word can provide the root of the word and its grammatical information. The

root word should be the most basic form of a word that is able to convey a particular description,

thought, or meaning. The definition given for root word is that it is a real word that can make

new words from root words by adding prefixes and suffixes [6]. Here all forms which are affixed

immediately after the root word is considered as suffix. No attempt was made to identify the


The Malayalam documents seen in the world wide web are collected and stores as Malayalam

corpus. Corpus [7] is a large collection of written and/or spoken text samples available in

machine-readable form, collected in a scientific way to represent the use of a language [8]. In this

work, a corpus of 24,000 words in written form is used for linguistic analysis of Malayalam from

different domains. Using the rule based approach, the rules that govern the suffixation are

derived manually by analysing the words in the corpus. The Root Word Identifier system will

separate the root words after removing the suffixes as per these rules.

Common grammatical categories for Malayalam are the noun, pronoun, verb, adverb, adjective,

postpositions, indeclinables, clitics etc. In this work, the main grammatical categories such as

noun, pronoun, verb, postpositions are analysed.

3.1. Noun Morphology

Nouns can occur in isolation or can take gender markers, plural markers, case suffixes,

postpositions, clitics etc. It takes the form

W= noun root ± [plural suffix] ± [case suffix] ± [postpositions] ± [clitics] ± …

where W is any word having the properties of a noun. Some of them are shown in Table 1.

International Journal of Computer Science & Information Technology (IJCSIT) Vol 4, No 3, June 2012


Table 1. Case markers in Malayalam with Example

Case Marker Example

Nominative nil mankan~

Accusative -e makane

Dative -kkU/(n)U makan

Sociative -ootU makanootU

Locative -il~ makanil~

Instrumental -aal~ makanaal~

Genitive -ute/nte makante

Plural forms of noun contain suffix '-maaR~' and 'kaL~'. But it also take some allomorphs 'ngngal'

and 'kkaL~'. Apart from these case markers and plural suffixes there are suffixes which are

agglutinated with nouns [9]. Some of them are allative marker (-ileekk), place locative suffix (-

athth), optative suffix (-aakatee), reciprocity suffix (-tammil), sufficient suffix(-maththi) etc.

Pronouns are those words which can be used instead of nouns. Since they are in free form,

considering them as root word for easy analysis: njaan~ (I (personal pronoun, first person,

singular)), nii (You (personal pronoun, second person, singular)), avan~ (He (Third person,

remote, singular, Masc.) etc. Sixty two pronouns are identified manually.

3.2. Verb Morphology

The morphological structure of verb is complex. They are capable of taking tense markers. Verbs

in Malayalam are not inflected for Person Number and Gender. All verbal forms in Malayalam

both finite and non finite consists of verb stems followed by affixes which express various

grammatical categories such as tense, aspect, mood, voice valency change [9][10][11] etc.

Generally Tense is classified into past, present and future. Aspect as perfective, imperfective,

progressive. Mood as indicative, interrogative, imperative, conditional, optative, debitive,

potential. Two voices are active voice and passive voice. Valency change is classified into

causative and passive.

Categorization of verbs [12] can be done according to the suffixes attached to verb forms.

Identified fifty-one verb suffixes to retrieve the verbal forms of words. Some of them are listed

below with suffixes and example

1. Past tense marker - /-njnju/, /-nnu/, /-RRu/, /-ththu/, /-thu/, /-i/, /-ccu/, /-Ntu/, /-ttu/

Eg: paRanjnju (told), ezhuthi (wrote)

2. Present tense marker - /-unnu/

Eg: varunnu (coming)

3. Debitive emphatic marker /-aNee/

Eg: tharaNee (should give)

International Journal of Computer Science & Information Technology (IJCSIT) Vol 4, No 3, June 2012


3.3. Postpositions

Robert Caldwell (1913), commented that every postpositions annexed to a noun constitutes,

properly speaking, a new case. On the basis of above definition, postpositions are also considered

here as markers. English language has prepositions instead of Malayalam postposition. Different

types of classifications are given for postpositions according to their origin, morphological

similarities such as particles, case indicators, co-ordinator, derived noun etc. Twenty-seven

suffixes are identified as postpositional suffixes. Some of them are

1. Duration/Distance suffix - /-oolaaM~/

Eg: raamanooLaaM~ (upto/till/ approximately raman)

2. Equality suffix - /-poole/

Eg: kittiyepoole (like a child)

3.4. Morphophonemic Changes in Malayalam

The suffixes obtained from the manual analysis are grouped according to their initial phonemes.

Group A contains all the suffixes starting with /-a/, Group AA will contain all the suffixes starting

with /-aa/ and so on. A detailed analysis of the morphophonemic change occurring when the final

syllable of the root word concatenate with initial phoneme of the suffix is carried out. They are

generalized and morphophonemic rules are derived to identify the root form of the word [13].

3.5. Morphophonemic Rule Implementation for Root Word Identification

A word in the corpus is analysed from right to left. The first encountered suffix part is removed

and if any link morph such as /in/, /u/ is present, they are also removed. The remaining part of the

word is taken for further analysis. If the last syllable is /y/ or /v/, it is sufficient to chop them for

obtaining the root form of the word. If the last syllable is /ththa/ convert this syllable to /-aM~/. In

this way, a list of rules is there to generate root form of words. Most of the words are confined to

these rules. Only limited exceptions are identified. Some of them are listed in Table 2. In these

words, some are having final syllables similar to suffix ending. They are stored separately as

exceptional words and considered them as root words.

Table 2. Example for list of exceptional words

Examples for list of exceptional word






The suffixes agglutinated with noun and verbs are classified as noun suffix, verb suffix and dual

functional suffix (suffixes which agglutinate with both). The NS, VS and DS as shown in Figure

1. contain these suffixes. These are the accessory files to the Root Word Identifier system. The

words from the corpus are fed to the Root Word Identifier system and morphophonemic rules as

International Journal of Computer Science & Information Technology (IJCSIT) Vol 4, No 3, June 2012


discussed in section 3.5 are applied to the word to obtain the root words. The system iteratively

identify the suffixes and word. At the end of the task, the root word is identified. The complete

details of the words from the Malayalam corpus are stored in Computational Lexicon. Block

diagram of the system is shown in Figure 1.

Figure 1. Block Diagram showing the input and output to Root Word Identifier

3.6. Algorithm for Root Word Identifier


step1: scan the word from right hand side

step2: identify the suffix S and compare S with VS

if present

store the word in verb category


step3: compare S in NS,DS and EX

if present

step4: Store the word according to the corresponding grammatical category

then, Remove S and apply morphophonemic rules

step5: repeat steps from 1 to 4 until the root word is encountered

step6: Store the root word in computational lexicon with grammatical category




Malayalam Words

Malayalam Corpus

Root Word


Computational Lexicon



International Journal of Computer Science & Information Technology (IJCSIT) Vol 4, No 3, June 2012



The work is implemented with Practical Extraction and Report Language (PERL) [14] in Linux

environment using Unicode supportive font.

Given a corpus as input, the Root Word Identifier system can generate a computational lexicon

having the linguistic details about the words in the corpus such as the root form of the word,

grammatical category of root word, inflected forms of the root word, all the suffixes agglutinated

with the word, name of each suffix, words obtained after removing each suffix and their

grammatical category.

An example which shows how the word kuttikaLutekuute meaning 'with the children' is processed

by the Root Word Identifier

[word from Malayalam corpus]

- + + +

[Phonetic notation]

kuttikaLutekuute- kutti + kaL~ + ute + kuute


word- child+ plural suffix + genitive case suffix + suffix showing interior movement

[Linguistic information]

kuttikaLutekuute-Noun word

kuttikaLute - Noun word

kuttikaL~ - Noun word

kutti - Noun word - Root word

An objective analysis of the performance of the system is done using the statistical measures -

Recall, Precision and F-measure.

1. Precision (P) = tp


2. Recall (R) = tp


3. F-measure= 2*P*R/(P+R)

where tp is the true positives, fp is the false positives and fn is the false negatives [15].

The typographical forms and foreign language words, which have less relevance in this work, are

removed from the test corpus and the remaining 23,045 words were used to test the system. Root

Word Identifier program will identify the words in the corpus and gives the root form of the word

with its grammatical category such as noun, verb or dual functional. The system showed

Precision, Recall and F-measure values as 95.42%, 95.05% and 95.22% when identifying root

words. Compound words and reduplicated words are stored as a single root word. Only a small

percentage of words were unprocessed and were processed wrongly. The system is having many

advantages when compared with existing morphological analysers. Only a monolingual

Malayalam corpus is required to develop a computational lexicon. So no need to have any lexicon

or machine readable dictionary or manual assistance for collecting words.

International Journal of Computer Science & Information Technology (IJCSIT) Vol 4, No 3, June 2012



As a part of developing a computational lexicon for Malayalam there arises a need for developing

a Root Word Identifier which identifies the root form of the word. A detailed morphological

analysis is carried out to study the underlying structure of Malayalam words. Morphophonemic

rules are derived to obtain the root form of the word automatically with its grammatical category

and inflected forms. This rule based approach with a larger corpus can contribute much to the

development of a full-fledged computational lexicon.