International Journal on Natural Language Computing (IJNLC) Vol. 5, No.

1, February 2016

A COMPREHENSIVE ANALYSIS OF STEMMERS

AVAILABLE FOR INDIC LANGUAGES
Harshali B. Patil, B. V. Pawar, and Ajay S. Patil

School of Computer Sciences, North Maharashtra University, Jalgaon, India

ABSTRACT
Stemming is the process of term conflation. It conflates all the word variants to a common form called as
stem. It plays significant role in numerous Natural Language Processing (NLP) applications like
morphological analysis, parsing, document summarization, text classification, part-of-speech tagging,
question-answering system, machine translation, word sense disambiguation, information retrieval (IR),
etc. Each of these tasks requires some pre-processing to be done. Stemming is one of the important building
blocks for all these applications. This paper, presents an overview of various stemming techniques,
evaluation criteria for stemmers and various existing stemmers for Indic languages.

KEYWORDS
Stemming, light weight, affix removal, HMM, n-gram, Indian languages

1. INTRODUCTION

In 21st century the e-data in regional languages appearing on internet increased drastically. But
due to the unavailability of NLP tools for these languages, access to this data is limited. The
scarcity of tools has attracted the attention of researchers and industry persons towards the
development of efficient systems for IR, text summarization, opinion mining, clustering,
classifications etc., for regional languages. The inflected nature of a language poses several
challenges for the automated processing of natural language data. In an information retrieval
system that does not use any word normalization, search results are directly affected as many
relevant documents are missed during the retrieval process. For instance, in case of the search
query “stemmer”, the system will not return documents containing “stemming” or “stem” if the
documents do not contain the term “stemmer”. Term conflation is the solution to these types of
problems. In many natural languages several words sharing the same morphological variant (root)
can be related to the same topic. Stemming is one of the popular term conflation techniques used
as a preliminary step in many natural language processing tasks. In linguistics stem is the form
that unifies the elements in a set of morphologically similar words. Stemming is the process
which determines the stem of the given word. The goal of a stemming algorithm is to reduce
variant word forms to a common morphological root, called stem [1]. India is a multilingual
country where there are 22 official languages which belongs to 4 different families of languages.
Among these 22 languages, 15 languages belong to Indo-Aryan, 4 are Dravidian, 2 are Tibeto-
Burman, and 1 belongs to Munda family.

This paper presents a comprehensive analysis of stemmers available for Indic languages. The
paper is organized as follows: related work is provided in section 2. Section 3 describes stemming
process in detail as well as existing stemmers for Indic languages and Section 4 concludes the
paper.
DOI: 10.5121/ijnlc.2016.5104 45
 International Journal on Natural Language Computing (IJNLC) Vol. 5, No.1, February 2016

2. RELATED WORK

Stemming is a well known research problem; but in early days of stemming it was studied only
for English language. Lovin’s stemmer is one of the oldest stemmer developed for English using
context sensitive longest match technique. Other most notable stemmers for English includes:
Porter’s stemmer, Dawson stemmer, Paice and Husk stemmer. Poter’s stemmer has became the
de-facto standard for English language. Some non-English stemmers were presented during 1990-
2000, but study related to stemming for Indic languages has been started after 2000. The
subsequent section discusses the work done related to stemmer development for Indic languages.
Most of the work related to stemmer development has most recently initiated. Sarkar et.al (2012)
surveyed stemmers for Bengali and has inferred that rule-based stemmers would be more suitable
for Bengali, whereas suffix strippers may not be sufficient for Bengali. The author also concludes
that POS tagged data and lexicon may improve performance in Bengali stemmer and accuracy
based evaluation techniques should be appropriate for measuring stemmer performance [21].
Lakshmi et. al. (2014) reviewed literature related to stemming algorithms for Indian and Non-
Indian languages and found that there is a need to develop a language independent stemmer for
all languages [27]. Madhurima et.al (2013) analyzed popular stemming algorithms supporting
information retrieval system and has shown that no perfect stemmer has been designed so far to
match all the requirements [22]. Kasthuri et. al (2014) comprehensively analyzed stemming
algorithms for Indian and non-Indian languages, but the analysis focuses on only recent stemmers
developed during 2010 to 2014 [8]. Bijal et. al overviewed stemming algorithms for Indian and
non-Indian languages and discussed 9 stemmers related to Indian languages [24]. Rakholia et. al.
(2014) presented comparative analysis of different stemmers and character recognition algorithms
for Indian Gujarati script where the author discussed the stemmers available for Gujarati language
only [25]. Sethi el.al (2014) presented a literature survey related to stemming algorithms for Odia
language [26]. Patil et.al (2014) presented a part-of-speech tagger for Marathi language using
limited training corpora and obtained 78.82% accuracy with the rule-based technique [41]. The
study related to developing links of compound sentences for parsing through Marathi link
grammar parser was carried out by Vaishali Patil et. al (2014) [42,43]. Nita Patil et.al (2016)
surveyed the name entity recognition (NER) systems with respect to Indian and foreign languages
and concludes that very less work on NER is reported for Indian languages like Marathi and
Guajarati [44]. Juhi et.al (2013) improved the quality of Gujarati-Hindi machine translation
through part-of-speech tagging and stemmer assisted transliteration and achieved 93.09% overall
efficiency of the transliteration scheme [45]. Kridanta analysis for Sanskrit has been done by
Murali et.al (2014) and achieved 92.75% and 95.37% precision and recall for Kridanta analyzer
[46].

3. STEMMING

The following section presents the detail information about stemming in terms of techniques used
to develop stemmers, various types of errors generated while stemming, evaluation criteria for
stemmers, and the existing stemmers for Indic languages. The major advantages and limitations
provided by stemming for information retrieval are as given below:

Advantages of stemming

• Use of stemmers in IR decreases index size because for all the terms that belongs to single
conflation class are reduced to single term. [16, 20].
• Reduction in index ultimately reduces the storage space required to store the inverted index
file. file [16, 47, 48].

46
 International Journal on Natural Language Computing (IJNLC) Vol. 5, No.1, February 2016

• Instead of using term variation if the stem is used for IR it increases the recall of retrieval
systems. [2, 20].
• It is used as an important component for pre-processing in many applications.

Limitations of stemming

• Exceptional cases are also grouped together e.g – university and universal – univers
• Sometimes decreases the retrieval performance

3.1. Classification of stemming techniques

The stemmers are broadly classified into two types: manual and automatic[48]. Manual stemmers
stems the terms manually while automated stemmers can be developed by using various
techniques. Simplest affix removal to a complicated technique like n-gram or Hidden Markov
Model (HMM) has been used for development of stemmers for various languages. Fig. 1
classifies the major stemming techniques used for stemmer development.

Figure 1. Some automatic stemming techniques

3.1.1. Affix removal Technique

These types of stemmers are also called as rule-based stemmers or suffix stripper. This is one of
the oldest and simplest techniques used for stemmer development. These types of algorithms uses
list of suffixes and with each suffix the criteria under which it may be removed from a word to
leave a valid stem. The affix removals based on rules are either done based on longest match
basis or in iterative manner. Fig 2 shows the procedure for affix removal stemmers. Some Indic
stemmers developed using this technique includes [3, 5, 9, 12, 38, 39].

Affix list Stripping

criteria

Term Stemming Stem

Figure 2. Affix removal stemming procedure

47
 International Journal on Natural Language Computing (IJNLC) Vol. 5, No.1, February 2016

The sample rule related to English affix removal / rule-based stemmer is as given below:

Rule :- IES -> I

E.g. :- Ponies -> Poni

3.1.2. Table / Dictionary lookup

In this technique a table of corresponding terms along with their stem is used. The stemming is
done by searching the corresponding term in the table and retrieving the stem related to that term
[48]. This technique is not popular for stemmer development due to limitations related to it: like
dictionary of term-stem are not available for many languages, the accuracy totally depends on the
size of dictionary, and dictionary are domain dependent; but this technique can be combined with
rule-based technique to develop hybrid stemmers and increase the stemming accuracy. Fig. 3
shows the process of stemming with help of dictionary / table lookup approach.

Corpus List of words

Documents to be
stemmed

Stemmed
Dictionary Stemming
words

Figure 3. Stemming procedure for dictionary / table lookup

3.1.3. Statistical Techniques

These types of stemmers uses some statistical measure for stemmer development like n-gram,
HMM, clustering-based method, corpus-based method. Some of them are briefly explained in the
following section. In this approach first the system is trained with the large corpus and then
inflected forms are submitted to the trained module for stemming.

N-gram:

N-gram is a set of n consecutive characters extracted from a word. In this technique the similar
words will have a high proportion of n-grams in common. The n-gram based stemmer calculates
the association measure between the pair of terms based on shared unique diagrams. Then
similarity measures are determined for all pairs of terms in the database, forming a similarity
matrix. After that terms are clustered using single link clustering. [40]

HMM:

The HMM is a finite sequence of states, and a set of transitions between states that are taken
based on the input observations. Each character comprising a word is considered as a state. All
possible states are divided into two groups (roots and suffixes) Word building process are defined
by the transitions between states. [29]

48
 International Journal on Natural Language Computing (IJNLC) Vol. 5, No.1, February 2016

Clustering-based:

In these types of stemmer first the Equivalence classes are discovered. Then a set of string
distance measures are defined. The lexicon for a given text collection is clustered using complete
linkage clustering technique to identify these equivalence classes. For some infrequent cases the
author uses some post-processing for each cluster. [2]

Corpus-based:

First the set of potential suffixes are formed. Then the equivalence classes are generated by
combining the common prefix and potential suffix information. Then the common prefix and
potential suffixes help to recognize the better equivalence classes through mutual agreement. [14]

3.1.4. Hybrid

When any two or more approaches are combined for stemmer development then the stemmer
becomes hybrid. Most hybrid stemmers were developed by combining lookup table with affix
removal. They overcome the disadvantages of one technique by combining it with other; for
example the lookup table based stemmer cannot able to stem if term is not present in the table so
when this type of stemmer is combine with affix removal then it can be able to stem the term
which are not present in dictionary. . The work related to hybrid stemmers for Indic languages has
been reported by [4, 6, 15, 29, 30, 37].

3.2. Comparison of stemming techniques

Table 1 compares the stemming techniques (rule based, dictionary lookup, statistical and hybrid)
in terms of advantages and limitations.

Table 1. Comparison of stemming techniques.

Technique Advantages Limitations

Rule-based Easy to implement, requires less memory, Rule preparation needs to be done
dictionary not required and manually, Over-stemming and under-
fast stemming error rate is high.
Dictionary has capability to work with exceptional Dictionary / table is not available,
lookup cases, under and over-stemming error rate storage and retrieval overhead,
is reduced and is fast accuracy depends on dictionary size.
Statistical Applicable for wide variety of languagesRequires large corpus to train the
Technique system, takes more processing time,
requires significant amount of memory.
Hybrid Combination of more than one technique, If dictionary lookup is combined then
overcomes the drawbacks of each other. extra overhead for storage.

3.3. Evaluation criteria for stemmers

Stemmer is used as a pre-processing component for various NLP applications. There are several
criteria that are used for evaluation of stemmers. Some of them include: correctness, retrieval
effectiveness, and compression performance [48]. Generally stemmer is evaluated based on
accuracy provided by it for stemming. Paice has given a method to evaluate stemmers based on
two types of error counting [33]. Under-stemming errors and over-stemming errors are used to
49
 International Journal on Natural Language Computing (IJNLC) Vol. 5, No.1, February 2016

evaluate the stemmer. Section 3.4 discusses these two types of errors. In IR few studies has
evaluated stemmers as a method for index compression while some discuss the improvement in
precision and recall for IR.

Frakes and Fox has given the stemmer strength metrics. [31,32] The degree to which a stemmer
changes words that it stems is called stemmer strength. Some ways to measure stemmer strength
are:
• The number of words per conflation class.
• Index compression factor.
• The number of words and stems that differ.
• The mean number of characters removed in forming stems.
• The median and mean modified Hamming distance between words ad their stems.

3.4. Types of errors in stemming

Generally two types of errors i.e over-stemming and under-stemming errors are generated while
stemming. [33]

3.4.1. Over-stemming

Over-stemming errors refers to the words that should not be grouped together by stemming but
they are grouped together. These types of errors will affect on precision of IR.

3.4.2. Under-stemming

Under-stemming refers to words that should be grouped together by stemming, but that are not
grouped together. These type of errors tend to decrease the recall in an IR search.

3.5. Existing stemmers for Indic languages

Significant work related to stemmer development has been done for non-Indic languages like
English, Arabic, etc. whereas stemming work is in progress for Indic languages. The work related
to stemmer development for various Indic languages has started after 2000. Till 2010, not much
work related to stemmer development has been reported for Indic languages. India is a
multilingual country where there are 22 official languages. Among these languages stemmers are
available for 13 Indic languages. Some language independent stemmers are also available. Table
2 presents the existing stemmers for Indic languages.

Table 2. Existing stemmers for Indic languages.

Sr. Author &

Language Year Approach Accuracy / Results
No. Reference
1 Results are favourable and
Ramanathan Suffix
2003 can be used effectively in
el al [9] removal
information retrieval
2 Pande et. al 89.90%
Hindi 2008 Unsupervised
[36]
3 Improvements over no
Dolamic Light and stemming: 19.3% with light
2010
et.al [11] aggressive stemmer, 27.6% with
aggressive stemmer
50
 International Journal on Natural Language Computing (IJNLC) Vol. 5, No.1, February 2016

4 Mishra et. al Hybrid 91.59%

2012
[37]
5 Gupta [5] 2014 Rule –based 83.65%
6 Sarkar et. al Rule-based 89% and above
2008
[39]
7 Zahurul et. Lightweight 90.80%
2008
al [10]
8 Bengali Improvements over no
Dolamic Light and stemming: 13.7% with light
2010
et.al [11] aggressive stemmer , 17.7% aggressive
stemmer
9 Das et. Al 0.4748 (Mean Average
2011 Rule-based
[12] Precision)
10 Kumar et. al Brute force 80.73%
2010
[13] algorithm
11 Punjabi Gupta et.al 87.37%
2011 Rule-based
[38]
12 Joshi et.al 95.6%
2014 Hybrid
[6]
13 Improvements over no
Dolamic et. Light and stemming: 13.9% with light
2010
al [11] aggressive stemmer , 41.6% with
Marathi
aggressive stemmer
14 Rule -based 80.7% for rule-based
Majgaonkar
2010 and 82.5% for unsupervised
et. al [3]
unsupervised
15 Ramchanrda Suffix 84.79%
2012
Tamil n et. al [19] stripping
16 Thangarasu Light 83.28%
2013
et. al [20] stemming
17 Chaupattnai Suffix 88%
Odia 2012
k et. al [17] stripping
18 Sethi [18] 2013 Light weight 85%
Hybrid (rule- 92%
Saharia et.
19 Assamese 2013 based and
al. [29]
HMM)
Patel el.al 67.86%
20 2010 Hybrid
[4]
Gujarati
Suba et.al Hybrid and 90.70% for hybrid
21 2011
[30] rule-based 70.70% for rule-based
Statistical 88.82%
22 Kannada Bhat [16] 2013
technique
80.02% for minimum suffix
Patra et. al
23 Kokborok 2012 Rule –based stripping, 85.13% for
[23]
maximum suffix stripping
Prajitha et. Suffix Computationally inexpensive
24 Malayalam 2013
al [35] stripping and domain independent.
Meitei et. 86.29%
25 Manipuri 2015 Hybrid
al.[15]
26 Telugu Kumar[34] 2013 Unsupervised 85.40%

51
 International Journal on Natural Language Computing (IJNLC) Vol. 5, No.1, February 2016

Gupta 86.50%
27 Urdu 2013 Rule-based
et.al[28]
Majumder Clustering - Improvement in the recall for
28 2007
et. al [2] based information retrieval
Statistical Outperformance observed
Paik et. al technique over the well known rule-
29 2011
[14] based stemmers for all
Language
languages under study
independent
Length-based Length based 79.63% (Urdu),
and 82.6% (Marathi), frequency
frequency- based 84.27% (Urdu), 63.5%
30 Husain [7] 2012
based (Marathi)
unsupervised
approach

Various techniques were used for the evaluation of stemmers as mentioned in section 3.3. It is
observed that the percentage accuracy provided by the stemmer was generally used to evaluate
the stemmers. Form the literature review related to Indic languages stemmer development it is
found that there is a difference in the accuracy levels provided by various types of stemmers.
Table 3 presents the percentage accuracy obtained by using various approaches for Indic
languages stemmer development.

Table 3. Accuracy ranges for stemming techniques

Accuracy range
Sr. No. Technique
(%)
1 Rule –based 80.02 – 89
2 Statistical 63.5 - 89.9
3 Hybrid 67.86 - 95.6

From table 3 it is observed that the maximum accuracy level provided by rule-based and
statistical techniques are nearly same but there is lots of difference between the minimum
accuracy levels achieved by both of these techniques. The maximum accuracy reported for Indic
languages are obtained by the use of hybrid technique.

4. CONCLUSION

Stemmer is an important and basic component in many natural language processing applications.
The accuracy of the stemmer strongly affects the results of the system in which it is used.
Various stemmer development techniques are being explored and studied for different languages
across the world. This paper surveyed the stemmers available for Indic languages. From literature
survey it is observed that the work related to stemmer development for some of the Indic
languages like Mizo, Santhali, etc has not been reported. The rule-based and light stemming
techniques are widely used for stemmer development for languages like Hindi, Marathi, Tamil,
etc. Hybrid approaches are also used by some researchers to avoid limitations of using single
technique. Hybrid technique has reported more accuracy than the other techniques for some of the
Indic languages. Though some work related to Indic stemmer development has been reported still
much work needs to be done. The techniques like dictionary lookup or hybrid approach needs to
be evaluated for the languages like Marathi, Hindi, etc. We intend to develop a stemmer for
Marathi language that will be used for efficient information retrieval. The survey of the existing
52
 International Journal on Natural Language Computing (IJNLC) Vol. 5, No.1, February 2016

techniques will be helpful in deciding the applicability of these techniques for Marathi stemmer
development.

ACKNOWLEDGEMENTS

The authors are thankful to University Grants Commission (UGC), New Delhi, India, for
financial assistance for the research study under the scheme of Special Assistance Programme
(SAP) of Departmental Research Support (DRS) Phase I.

REFERENCES

[1] M. Bacchin, N. Ferro, & M. Melucci (2002) “The Effectiveness of a Graph-based Algorithm for
Stemming”, E.-P. Lim et. al (Eds.): ICADL 2002, LNCS, 2555, pp 117-128.
[2] P. Majumder, M. Mitra, S. K. Parui, and G. Kole , P. Mitra and K. Datta, (2007) “YASS: Yet Another
Suffix Stripper”, ACM Transactions on Information Systems, Vol. 25, No. 4, Article 18.
[3] M. M. Majgaonkar & T. J. Siddiqui, (2010) “Discovering Suffixes: A Case Study for Marathi
Language”, Int. Journal on Computer Science and Engineering Vol. 02, No.08, pp 2716-2720.
[4] P. Patel, K. Popat & P. Bhattacharyya, (2010) “Hybrid Stemmer for Gujarati”, Proc. of the 1st
workshop on South & Southeast Asian Natural Language Processing, pp 51-55.
[5] V. Gupta, (2014) “Hindi Rule-based Stemmer for Nouns”, International Journal of Advanced
Research in Computer Science and Software Engineering, Vol. 4, No.1, pp 62-65.
[6] G. Joshi & K. D. Garg, (2014) “ Enhanced Version of Punjabi Stemmer using Synset”, Int. Journal of
Advanced Research in Computer Science and Software Engineering, Vol. 4, No.5, pp 1060-1065.
[7] M. S. Husain, (2012) “An Unsupervised Approach to Develop Stemmer”, Int. Journal on Natural
Language Computing Vol. 1, No.2, pp 15-23.
[8] M. Kasthuri & S. B. R. Kumar, (2014) “A Comprehensive Analysis of Stemming Algorithms for
Indian and non-Indian languages”, Int. Journal of Computer Engineering and Applications, Vol. 7,
No. 3, pp 1-8.
[9] A. Ramanathan, D. D Rao, (2003) “A Lightweight Stemmer for Hindi”, Proc. of the EACL.
[10] M. D. Zahurul Islam, M.D.N. Uddin & M. Khan, (2007) “A Light Weight Stemmer for Bengali & Its
Use in Spelling Checker”, Proc. of Ist Int. Conference on Digital Communications & Computer
Applications.
[11] L. Dolamic, J. Savvo, (2010) “Comparative Study of Indexing and Search Strategies for the Hindi,
Marathi and Bengali Language”, Special Issue of ACM Transaction on Asian Language Information
Processing on IR for Indian Languages Vol. 9 No. 3.
[12] S. Das, P. Mitra, (2011) “A Rule Based Approach of Stemming for Inflectional and Derivational
Words in Bengali”, Students' Technology Symposium IEEE No. Date: 14-16 Jan. 2011 pp 134 - 136
Kharagpur.
[13] D. Kumar, P. Rana, (2011) “Stemming of Punjabi Words by Using Brute Force Technique”, Int.
Journal of Engineering Science and Technology Vol. 3 No. 2 pp. 1351-1358.
[14] J.H. Paik, & S. K. Parui, (2011) “A Fast Corpus Based Stemmer”, ACM Transactions on Asian
Languages Information Processing Vol. 10, No. 2, Article No. 2.
[15] S. P. Meitei, B. S. Purkayastha & H. M. Devi, (2015) “Development of a Manipuri Stemmer: A
Hybrid Approach”, Int. Symposium on Advanced Computing & Communication 2015.
[16] S. Bhat, (2013) “Statistical Stemming for Kannada”, The 4th Workshop on South and Southeast
Asian NLP, Int. Joint Conference on Natural Language Processing, pp 25-33.
[17] S. Chaupattnaik, S. S. Nanda, S. Mohanty, (2012) “A Suffix Stripping Algorithm for Odia Stemmer”,
Int. Journal of Computational Linguistics and Natural Language Processing.
[18] D. P. Sethi, (2013) “Design of Lightweight stemmer for Odia Derivational Suffixes”, Int. Journal of
Advanced Research in Computer and Communication Engineering, Vol. 2,No. 12.
[19] V. A. Ramchandran and I Krishnamurthi, (2012) “ An Iterative Suffix Stripping Tamil Stemmer”,
Proceedings of the International Conference on Information Systems Design and Intelligent
Applications, Advances in intelligent and Soft Computing, Vol 132. pp 583-590.
[20] M. Thangarasu & R. Manavalan, (2013) “Stemmers for Tamil language: Performance Analysis”, M.
Thangarasu et. al./ Int. Journal of Computer Science & Engineering Technology, Vol. 4, No. 7.
53
 International Journal on Natural Language Computing (IJNLC) Vol. 5, No.1, February 2016

[21] S. Sarkar, & S. Bandyopadhyay, (2012) “On the Evolution of Stemmers: A Study in the Context of
Bengali Language”, Int. Journal of Computational Linguistics and Natural Language Processing, Vol.
1, No. 2, pp 51-59.
[22] Madhurima V., T. V. N. Rao, & L. S. Bhargavi, (2013) “ Analysis of Popular Stemming Algorithms
Supporting Information Retrieval system”, Int. Journal of Computer Organization Trends, Vol. 3, No.
9, pp 377-385.
[23] B. G. Patra, K. Debbarma, S. Debbarma, D. Das, A. Das, S. Bandyopadhyay (2012) “A Light Weight
Stemmer in Kokborok”, Proc. of the 24th Conference on Computational Linguistics and Speech
Processing, pp -318-325.
[24] D. Bijal, & S. Sanket, (2014) “Overview of Stemming Algorithms for Indian and Non-Indian
Languages”, Int. Journal of Computer Science and Information Technologies, Vol. 5, No. 2, pp 1114-
1146.
[25] R. M. Rakholia & J. R. Saini, (2014) “A Study and Comparative analysis of Different Stemmers and
Character Recognition Algorithms for Indian Gujarati Script”, Int. Journal of Computer Applications,
Vol. 106, No. 2, pp 45- 50.
[26] D. P. Sethi, & S. K. Barik, (2014) “A Literature Survey: Stemming Algorithm for Odia Language”,
Int. Journal of Advanced Research in Computer Engineering & Technology, Vol. 3. No. 1 pp 9- 11.
[27] R. V. Lakshmi, & S. B.R. Kumar, (2014) “ Literature Review: Stemming Algorithms for Indian and
Non-Indian Languages”, Int. Journal of Advanced Research in Computer Science & Technology, Vol.
2, No. 3, pp. 349-352.
[28] Gupta V., Joshi N., Mathur I. (2013), “Rule Based Stemmer in Urdu”, Proc. of 4th Int. Conference on
Computer and communication Technology, pp 129-132.
[29] N. Saharia, K. M. Konwar, U. Sharma, J. K. Kalita, (2013) “An Improved Stemming Approach using
HMM for a Highly Inflectional Language”, A. Gelbukh (Ed). CICLing 2013, part I, LNCS 7816, pp
164-173.
[30] K. Suba, D. Jiandani, & P. Bhattacharyya (2011) “Hybrid Inflectional Stemmer and Rule-based
Derivational Stemmer for Gujarati”, Proc. of 2nd Workshop on south & Southeast Asian Natural
Language Processing, pp 1-8.
[31] S. R. Sirsat, V. Chavan, & H. S. Mahalle (2013), “Strength and Accuracy Analysis of Affix Removal
Stemming Algorithms”, Int. Journal of Computer Science and Information Technologies, Vol. 4,
No.2, pp 265-269.
[32] Frakes W. B., & Fox C. J. (2003) “Strength and Similarity of Affix Removal Stemming Algorithms”,
ACM SIGIR Forum, Vol. 37, No.1, pp 26-30.
[33] Paice C. D. (1996) “Method for Evaluation of Stemming Algorithms Based on Error Counting”,
JASIS, Vol. 47, No.8, pp 632-649.
[34] A. P. Siva Kumar, Dr. P. Premchand, & Dr. A. Govardhanv (2011) “TelStem:An Unsupervised
Telugu Stemmer with Heuristic Improvements and Normalized Signatures”. Int. Journal of
Computational Linguistics and Applications ,Vol. 2, No. 1 , pp 13-23.
[35] Prajitha U., Sreejith C, P. C. R. Raj (2013) “LALITHA: A Light Weight Malayalam Stemmer using
Suffix Stripping Method”, Int. Conference on Control Communication and Computing, pp 244-248.
[36] A. K. Pandey ,T. J Siddiqui, (2008) “An Unsupervised Hindi Stemmer with Heuristic Improvements”,
Proc. of the Second Workshop on Analytics for Noisy Unstructured Text Data ACM, pp. 99-105.
[37] U. Mishra, C. Prakash (2012) “MAULIK: An Effective Stemmer for Hindi Language” Int. Journal on
Computer Science and Engineering Vol. 4. No. 5 pp. 711 – 717
[38] V. Gupta, G. S. Lehal (2011) “Punjabi Language Stemmer for Nouns and Proper Names”, Proc. of
the 2nd Workshop on South and Southeast Asian Natural Language Processing, pp. 35-39.
[39] S. Sarkar, S. Bandyopadhyay (2008) “Design of a Rule-based Stemmer for Natural Language Text in
Bengali”, Proc. of the IJCNLP – 08 Workshop on NLP for Less Privileged Languages, pp 65 -72.
[40] J. Mayfield ,P. Mcnamee, (2003) “Single N-Gram Stemming”, Proc. of the 26th Annual Int. ACM
SIGIR Conference on Research and Development in Information Retrieval.
[41] H. B. Patil, A. S. Patil and B. V. Pawar (2014) “ Part-of-Speech Tagger for Marathi Language using
Limited Training Corpora”, IJCA Proc. on National Conference on Recent Advances in Information
Technology NCRAIT(4) , pp 33-37.
[42] Vaishali. B. Patil & B. V. Pawar (2015) “Modeling Complex Sentences for parsing through Marathi
Link Grammar Parser”, Int. Journal of Computer Science Issues, Vol. 12, Iss. 1, No 2, pp 108-113.

54
 International Journal on Natural Language Computing (IJNLC) Vol. 5, No.1, February 2016

[43] Vaishali. B. Patil & B. V. Pawar (2014) “Developing Links of Compound Sentences for Parsing
Through Marathi Link Grammar”, Int. Journal on Natural Language Computing, Vol. 3, No 5/6, pp
108-113.
[44] Nita Patil, Ajay S. Patil & B. V. Pawar (2016) “Survey of Named Entity Recognition Systems with
respect to Indian and Foreign Languages”, Int. Journal of Computer Applications, Vol. 134, No. 16,
pp 21-26.
[45] Juhi Ameta, Nisheeth Joshi & Iti Mathur (2013) “Improving the Quality of Gujarati-Hindi Machine
Translation through Part-of-Speech Tagging and Stemmer Assisted Transliteration”, Int. Journal on
Natural Language Computing, Vol. 2, No.3, pp 49-54.
[46] N. Murali, Dr. R.J. Ramasreee & Dr. K.V.R.K. Acharyulu (2014) “Kridanta Analysis for Sanskrit”,
Int. Journal on Natural Language Computing, Vol. 3, No.3, pp 33-49.
[47] C. Moral, A. Antonio, R. Imbert & J. Ramírez (2014) “A Survey of Stemming Algorithms in
Information Retrieval”, Information Research, Vol. 19, No. 1.
[48] William B. Frakes and Ricardo Baeza-Yates (1992) Information Retrieval: Data Structures &
Algorithms, Prentice Hall

55