ISSN 2348-1218 (print)
International Journal of Interdisciplinary Research and Innovations ISSN 2348-1226 (online)
Vol. 6, Issue 3, pp: (142-147), Month: July - September 2018, Available at: [Link]
Grammar Checking Using Pattern Matching
Approach
1
Leekha Jindal, 2Vijay Rana
1
Research Scholar, Sant Baba Bhag Singh University, Jalandhar
2
Assistant Professor, Department of Computer Science and Applications, Sant Baba Bhag Singh University, Jalandhar
Abstract: In this research articles author proposed a pattern matching approach for grammar checking of Punjabi
sentences. In the proposed approach author used the pattern of sentences whose length ranges from 4 to 6. An
annotated corpus of around 10,000 sentences is used for generating the patterns. Different patterns are generated
for sentences of different lengths.
Keywords: grammar checker, pattern matching approach.
1. INTRODUCTION
Grammar checker also called syntactic analyzer is a software that verify the syntax of a specific language against the
grammatical rules of that language. If the written text is according to the rule of the language then no error will be
produced by the grammar checker otherwise if the written text is not according to the grammar of the language then it will
generate an error message and will also display a list of suggestions to rectify these errors. Generally grammar checking
systems are of not used independently rather these are part of some word processor like MS WORD. Many grammar
checkers have been developed for English and other languages. Very little work has been done for Indian languages.
Probably, Bangla grammar checker [19], Urdu [21] and Punjabi grammar checker [16] [26][27][28] are the only systems
developed for Indian languages.
2. EXISTING WORK
Various techniques have been used by various authors for different languages [22][23][24][25]. Some of these includes
Schmaltz, A. [Link] (2017) [1] proposed seq2seq approach for error detection and correction in sentence that achieved a F0.5
score as 50.12 for deletion method, 42.51 for insertion method and 50.39 for replacement method as compare to SMT
approach that has F0.5 score 46.56 for deletion, 31.48 for insertion and 42.21 for replacement methods when tested on
Automated Evaluation of Scientific Writing dataset. Sharma, S.K. [Link] (2016) [2] put their efforts to improve the existing
rule based Punjabi grammar checker and observed 5-6% improvement in morph and 8-9% improvement in POS tagger.
Schmaltz, A. [Link] (2016) [3] proposed an attention based encoder-decoder model for checking grammar errors present in
a sentence and showed precision as 0.5444, recall as 0.7413 and F-score as 0.6278. Lin, C. J. [Link] (2015) [4] proposed a
system that checked the grammatical accuracy of Chinese sentences generated by deleting, inserting or exchanging
characters or words and achieved a precision of 23.4% and a recall of 36.4% in the identification level. Boroș, T [Link]
(2014) [5] described the development of RACAI’s (Research Institute for Artificial Intelligence) hybrid grammatical error
detection and correction system. Temesgen, A. [Link] (2013) [6] described the statistical grammar checker for Amharic
language. This system showed 59.72% precision and 82.69% recall for simple sentences using bi-gram and if trigram is
used then system showed 67.14% precision and 90.38% recall. Further for complex sentences system shows 57.82%
precision and 65.38% recall for bigram and 63.76% precision and 67.69% recall for trigram. Xing, J. W. [Link] (2013)[7]
described the hybrid approach based NLP-CT Grammatical Error Detection and Correction system for the CoNLL 2013
shared task. Nazar, R. [Link] (2012)[8] explored the possibility of using a large n-gram corpus (Google Books) to derive
lexical transition probabilities from the frequency of word n-grams and then use them to check and suggest corrections in
a target text without the need for grammar rules and obtained a precision of 64.58, Recall 47.69 and F measure 54.86.
Sharma, S.K. [Link] (2011) [9] used Hidden Markov Model (HMM) to improve the accuracy of existing rule based Punjabi
POS tagger. Jiang, Y. et al. (2011) [10] proposed a rule based grammar checker. Tesfaye, D. (2011) [11] developed a rule
Page | 142
Research Publish Journals
ISSN 2348-1218 (print)
International Journal of Interdisciplinary Research and Innovations ISSN 2348-1226 (online)
Vol. 6, Issue 3, pp: (142-147), Month: July - September 2018, Available at: [Link]
based grammar checker for Afan Oromo (language widely spoken and used in Ethiopia). Kasbon, R. (2011) [12] proposed
a rule based grammar checking system for Malay language. Deksne, D. [Link] (2011) [13] explained the implementation of
the Latvian grammar checker based on a parser. Henrich, V [Link] (2009) [14] proposed a Language Independent Statistical
Grammar (LISG) checking system. They developed this system by using N-gram statistical technique. Singh M [Link]
(2008)[15] developed a rule based grammar checker for Punjabi that reported precision of 76.79%, recall of 87.08%, and
F-measure of 81.61%. Kumar, A. [Link] (2007) [16] provided a corpus based approach for grammar checking that is based
on the principles of an Artificial Immune System (AIS). Bal, B. K. [Link] (2007) [17] proposed an architecture of rule based
grammar checking system for Nepali language. Alam, M. Jahangir [Link] (2006)[18] used Part of speech (POS) tags and n-
gram technique to check the grammatical correctness of a sentence. The system showed an accuracy of 63% for English
and 53.70% for Bangla when tested on manually tagged correct sentences. Sjöbergh, J. [Link] (2005) [19] describes a
method to create an automated grammar checker that did not require any manual work. Kabir, H. [Link] (2002) [20]
proposed a computational model for Urdu Grammar Checker. Two pass parsing approach was used for analysis of
sentence in which first, some base Phrase Structure Grammar (PSG) Rules are used to parse the sentence and in case of
failure, Movement Rules are applied and sentence is reparsed. Arppe, A. [Link] (1999) [21] developed a commercial
grammar checking system for Swedish at Lingsoft Inc. This grammar checker is also a part of the Microsoft Office since
2000.
3. PROPOSED MODEL
The proposed architecture for classification of sentences on the basis of the length is shown in figure 1.
Figure 1: Proposed architecture for classification of sentences
Page | 143
Research Publish Journals
ISSN 2348-1218 (print)
International Journal of Interdisciplinary Research and Innovations ISSN 2348-1226 (online)
Vol. 6, Issue 3, pp: (142-147), Month: July - September 2018, Available at: [Link]
As shown in figure1, after inputting the corpus, the corpus is broken into individual sentences. Each individual sentence
undergo pre-processing in which spell checking is performed so as to make the corpus free from spelling errors. Other
than spell checking, duplicate entries of sentences as well as incomplete sentences are removed from the corpus. After
pre-processing, the corpus is classified into various classes according to size of sentence. For classifying the sentences,
each individual sentence is inspected for its length and separate class is formed for each specific length. In this research
author experimented with three lengths i.e. from length 5 to length 7. Sentences having length less than 5 and greater than
7 are discarded.
Input Sentence
Tokenization
Apply Morph
Contains
Unknown
words
Apply POS tagger
Contains
Ambiguous Discard
tags
Create Patterns
Testing Trainings
Figure 2: Proposed algorithm for pattern generation
Now as shown in figure2, after classification, each word in every sentence present in each class is assigned its
grammatical information in the form of part of speech (POS) tags. Morphological analyzer is used for this task. After
applying morphological analyzer (MA), sentences containing unknown tags i.e. those sentences in which one or more
words are assigned an unknown tag are removed from the list. After that HMM based POS tagger is applied to keep only
appropriate tag with each word out of multiple tags assigned by MA. Now we left with sentences having word tag
combinations. After this only tag pattern is extracted from word tag pairs. These tag patterns are divided in to two parts
one part is used for training and second part is used for testing.
Page | 144
Research Publish Journals
ISSN 2348-1218 (print)
International Journal of Interdisciplinary Research and Innovations ISSN 2348-1226 (online)
Vol. 6, Issue 3, pp: (142-147), Month: July - September 2018, Available at: [Link]
Input Sentence
Tokenization
Apply Morph
Contains
Unknown
words
Apply POS tagger
Contains
Ambiguous Grammar cannot be checked
tags
Create Patterns
Not
Matched
Match
Grammatically Patterns
with the
In-correct generated
pattern from training
corpus
Matched
Grammatically
Correct
Figure 3: Grammar checking using pattern matching
The grammar checking process is shown in figure 3. Input sentence is first passed through morph and if any of the word
in the sentence is marked unknown then the grammar of that sentence cannot be checked. If all the words of the sentence
are tagged then POS tagger is applied to remove the ambiguity of the tags if any present. If even after applying the POS
tagger some word still contains ambiguous tags then also the grammar of this sentence cannot be checked. And if the
sentence does not contain unknown word as well as ambiguous tag then pattern of the tags is extracted from the sentence
and this pattern is searched in the existing database of patterns generated during training. If the database contains the
pattern then the sentence is grammatically correct otherwise sentence is not grammatically correct.
Page | 145
Research Publish Journals
ISSN 2348-1218 (print)
International Journal of Interdisciplinary Research and Innovations ISSN 2348-1226 (online)
Vol. 6, Issue 3, pp: (142-147), Month: July - September 2018, Available at: [Link]
4. RESULTS AND DISCUSSION
Total Number of Number of sentences Number of sentences Number of sentences Number of sentences
sentences in corpus having length 5 having length 6 having length 7 having length more
than 7 (Discarded)
10220 2152 4317 2511 1240
Number of sentences having length 5 Number of sentences having length 6 Number of sentences having length 7
2152 4317 2511
Used for Training Used for testing Used for Training Used for testing Used for Training Used for testing
1152 1000 3317 1000 1511 1000
5. CONCLUSION AND FUTURE SCOPE
This grammar checker can be helpful in development of other natural language processing systems like question
answering, dialogue generation, paraphrasing and machine translation systems etc. Further the statistical techniques used
proposed in this research article can also be implemented in other Indian languages having similar features as that of
Punjabi. Also after implementing the statistical approach, hybrid approach can also be used for further enhancement of the
system.
REFERENCES
[1] Schmaltz, A., Kim, Y., Rush, A. M., & Shieber, S. M. (2017). Adapting sequence models for sentence
correction. arXiv preprint arXiv:1707.09067.
[2] Sharma, S. K., & Lehal, G. S. (2016, March). Improving Existing Punjabi Grammar Checker. In Computational
Techniques in Information and Communication Technologies (ICCTICT), 2016 International Conference on (pp.
445-449). IEEE.
[3] Schmaltz, A., Kim, Y., Rush, A. M., & Shieber, S. M. (2016). Sentence-level grammatical error identification as
sequence-to-sequence correction. arXiv preprint arXiv:1604.04677.
[4] Lin, C. J., & Chen, S. H. (2015, July). NTOU Chinese Grammar Checker for CGED Shared Task. In Proceedings of
The 2nd Workshop on Natural Language Processing Techniques for Educational Applications (pp. 15-19).
[5] Boroș, T., Dumitrescu, S. D., Zafiu, A., Tufiș, D., Barbu, V. M., & Văduva, P. I. (2014). RACAI GEC–A hybrid
approach to Grammatical Error Correction. CoNLL-2014, 43.
[6] Temesgen, A., & Assabie, Y. (2013). Development of Amharic Grammar Checker Using Morphological Features of
Words and N-Gram Based Probabilistic Methods. IWPT-2013, 106.
[7] Xing, J. W., Wang, L. Y., Wong, F., Chao, S., & Zeng, X. D. (2013, August). UM-Checker: A hybrid system for
English grammatical error correction. In The seventeenth conference on computational natural language learning:
shared task, Aug. 2013, p. 34-42.
[8] Nazar, R., & Renau, I. (2012, April). Google books n-gram corpus used as a grammar checker. In Proceedings of the
Second Workshop on Computational Linguistics and Writing (CLW 2012): Linguistic and Cognitive Aspects of
Document Creation and Document Engineering (pp. 27-34). Association for Computational Linguistics.
[9] Sharma, S. K., & Lehal, G. S. (2011, June). Using hidden markov model to improve the accuracy of punjabi pos
tagger. In Computer Science and Automation Engineering (CSAE), 2011 IEEE International Conference on (Vol. 2,
pp. 697-701). IEEE.
[10] Jiang, Y., Wang, T., Lin, T., Wang, F., Cheng, W., Liu, X., & Zhang, W. (2012, June). A rule based Chinese
spelling and grammar detection system utility. In System Science and Engineering (ICSSE), 2012 International
Conference on (pp. 437-440). IEEE.
[11] Tesfaye, D. (2011). A rule-based Afan Oromo Grammar Checker. IJACSA Editorial.
[12] Kasbon, R., Amran, N. A., Mazlan, E. M., & Mahamad, S. (2011). Malay language sentence checker.
Page | 146
Research Publish Journals
ISSN 2348-1218 (print)
International Journal of Interdisciplinary Research and Innovations ISSN 2348-1226 (online)
Vol. 6, Issue 3, pp: (142-147), Month: July - September 2018, Available at: [Link]
[13] Deksne, D., & Skadiņš, R. (2011). CFG Based Grammar Checker for Latvian. NODALIDA 2011 Conference
Proceedings, pp. 275–278.
[14] Henrich, V., & Reuter, T. (2009). LISGrammarChecker: Language Independent Statistical Grammar
Checking. Hochschule Darmstadt & Reykjavík University.
[15] Gill, M. S., & Lehal, G. S. (2008, August). A grammar checking system for Punjabi. In 22nd International
Conference on Computational Linguistics: Demonstration Papers (pp. 149-152). Association for Computational
Linguistics.
[16] Kumar, A., & Nair, S. (2007). An artificial immune system based approach for English grammar checking. Artificial
immune systems, 348-357.
[17] Bal, B. K., Shrestha, P., Pustakalaya, M. P., & PatanDhoka, N. (2007). Architectural and System Design of the
Nepali Grammar Checker. PAN Localization Working Paper.
[18] Alam, M. Jahangir, Naushad UzZaman, and Mumit Khan. 2006. N-gram based Statistical Grammar Checker for
Bangla and English. In Proc. of ninth International Conference on Computer and Information Technology (ICCIT
2006).
[19] Sjöbergh, J., & Knutsson, O. (2005, September). Faking errors to avoid making errors: Very weakly supervised
learning for error detection in writing. In Proceedings of RANLP (pp. 506-512).
[20] Kabir, H., Nayyer, S., Zaman, J., & Hussain, S. (2002, December). Two pass parsing implementation for an Urdu
grammar checker. In Proceedings of IEEE international multi topic conference (pp. 1-8).
[21] Arppe, A. 2000. Developing a grammar checker for Swedish. In The 12th Nordic Conference of Computational
Linguistics pp. 13-27.
[22] Blossom Manchanda, Vijay Anant Athavale and S K Sharma, Various Techniques Used for Grammar Checking.
International Journal of Computer Applications & Information Technology, 177-181, Special Issue on NLP, June
2016.
[23] Misha Mittal, Dinesh Kumar, S K Sharma, Grammar Checker for Asian Languages: A Survey. International Journal
of Computer Applications & Information Technology, 163-167, Special Issue on NLP, June 2016
[24] Sharma, S. K. Role of Statistical Approaches for Error Detection in English and Other European Languages. AGU
International Journal of Research in Social Sciences & Humanities. 234-239, Vol. 5, July Dec-2017.
[25] Sharma, S. K Rule Based Grammar Checking Systems (A Survey). International Journal of Computer Applications
and Information Technology.217-220, 10(1) July 2016.
[26] Sanjeev Kumar Sharma, Effect of Statistical POS tagger on Syntactic Analysis of Punjabi Sentences. Indian Journal
of Science and Technology, Vol9 (32), August 2016
[27] S.K. Sharma, G.S. Lehal, Improving Existing Punjabi Grammar Checker. IEEE International Conference on
Computation Techniques in Information and Communication Technologies held at Indraprastha University, New
Delhi, IEEE Xplore, 2016
[28] Sharma, S. K. Detection and Correction of Style Errors Present in Punjabi Sentences. International Journal of
Computer Science and Engineering, 196-199, 5(8), August 2017
Page | 147
Research Publish Journals