# Introduction atural Language Processing(NLP) is a theoretically motivated multiple methods and techniques from which are selected for the accomplishment of particular type of language in analyzing and representing a human communicable at one or more level of linguistic analysis in the purpose of achieving human like languages processing for a range of tasks or applications. Word Sense Disambiguation (WSD) [2] is the process of differentiating among the senses of words. The process of selecting most appropriate meaning of the word based on the context in which they occur. Computational identification of meaning for words in context is called Word Sense Disambiguation. WSD [3] process to remove the ambiguity of word in a given context is an important for NLP applications such as Information Retrieval, Machine # Approach for Two Word Disambiguation Two Word Disambiguation Rules Morphological analysis [10], [13] of a word gives detailed information about a word. Morphologically [11] every word carries information with reference to its lexemic form, morpho syntactic [12] category, and inflection. The detailed information may include among many other features, such as root/stem i.e. the lexemic shape listed in the dictionary the lexical category like noun/verb/adjective/adverb/pronoun /number /indeclinable as the case may be. The following are some of the POS tag [4], [5] [6] disambiguation rules [7], [8], [9] used in the task: W1 :: W2 => W1 :: W2 Where W1 and W2 a sequence of words in that order. Where n is noun, v is verb, pn is pronoun, adj is adjective and adv is adverb. Here from rule 2 when a word carries tags (n,pn) and followed by another word carrying the tag n then the tag pn retained eliminating the n from (n,pn). From rule 10 a word carrying the tag such as (n,pn) followed by avy then most the times pn will be retained and v will be eliminated. Depending on the context linguist will decide which tag will be retained and which one has to be eliminated. These are mostly contextually based syntactic rules. If two word sequences is unable to resolved unique tags then three words, four words sequence rules may be used for disambiguation. # III. Theoritical Explanation with Example for Two Word Ambiguity Here in the above sentence the word carries tags (n,adj) and followed by another word carrying the tag n then the tag adj retained eliminating the n from (n,adj).so from the above sentence adj is eliminated and n is retained. # c) After Applying Disambiguation Rule Adaxi a Nacivewaku alavAtu padipoyiMxi . n n n v punc Where punc is punctuation. # d) Analysis of Two Word Disambiguation Here the below figures 1 and 2 explores the analysis of the Accuracy. Where X-axis indicates the number of test sessions and Y-axis indicates the Accuracy. As the result, we found that the proposal method can disambiguate nearly 98%. :: w2 :: w3 => w1 :: w2 :: w3 n,v,pn :: n :: pn,v => v :: n :: pn In the above sentence the first word carries tags (n,v,pn) and followed by second word carrying the tag n and followed by third word carrying the tags (pn,v) then the tag v retained from the first word and pn retained from the third word eliminating the (n,pn) from (n,v,pn) and eliminating v from (pn,v). iv. Analysis Of Three Word Disambiguation Here the above figures 3 and 4 explores the analysis of the Accuracy. Where X-axis indicates the number of test sessions and Y-axis indicates the Accuracy. As the result, we found that the proposal method can disambiguate nearly 96%. We are very thankful to all the authors in a reference list, to make this research article in a better shape and right direction. # Conclusion and Future Research Direction This research article explores the impact of twoword disambiguation and three-word disambiguation. , ![Global Journal of Computer Science and TechnologyVolume XIV Issue I Version I Journals Inc. (US)S.NO SENTENCE ID BEFORE DISAMBIGUATION RULE AFTER DISAMBIGUATION RULE (RESULT)](image-2.png "N") 1![Figure 1 : Two word disambiguation rules accuracy](image-3.png "Figure 1 :") 2![Figure 2 : Two word disambiguation rules accuracy](image-4.png "Figure 2 :") 3![Figure 3 : Three word disambiguation rules accuracy](image-5.png "Figure 3 :") 4![Figure 4 : Three word disambiguation rules accuracy v.](image-6.png "Figure 4 :") ![](image-7.png "") 28926n :: v,n :: v,pn=>n :: n :: vn :: n :: v911634n,v : avy :: v,pn,adj =>n :: avy :vn :: avy :v © 2014 Global Journals Inc. (US) Disambiguation and Empirical approach for Three-Word WSD Here based on the context, linguist will decide which tag will be retained and which one has to be eliminated. We observed that if two-word and three-word sequences is unable to resolve unique tags, then four-word, five-word sequence rules may be useful for disambiguation. * The swarthmore college senseval-3 system RichardWicentowski EmilyThomforde AdrianPackel Proceedings of Senseval-3, Third International Workshop on Evaluating Word Sense Disambiguation Systems Senseval-3, Third International Workshop on Evaluating Word Sense Disambiguation Systems 2004 * Introduction to the Special Issue on Word Sense Disambiguation: The State of the Art Nancy; JeanIde Véronis Computational Linguistics 24 1 1998 * The Interaction of Knowledge Sources in Word Sense Disambiguation Mark; YorickStevenson Wilks Computational Linguistics 27 3 2001 * Part of Speech Tagging LYHalevi Seminar in Natural Language Processing and Computational Linguistics Israel April, 2006 School of Computer Science, TeL Aviv University * SPEECH and LANGUAGE PROCESSING: An Introduction to Natural Language Processing, Computational Linguistic and Speech Recognition DJurafsky JHMartin Pattern Recognition, Image Analysis and Applications: Lecture Notes in Computer Science Prentice-Hall 2000. 2004. 2004 3287 * Applied morphological processing of English GMinnen JCarroll DPearce Natural Language Engineering 2001 Cambridge University Press * Disambiguation Rules for Telugu Language Sentences: A Practical Approach * chapter 8: Word classes and Part of Speech Tagging DJurafsky JHMartin 10.1109/ARTCom.184 Speech and Language Processing IEEE Press 2000. 2009 * Why is there Morphology DieterWunderlich 23th Annual Meeting of the DGIS 2-4, 2004 12 * Disambiguation in Myanmar Word Segmentation WPPa NLThein Proceedings Of the Seventh International Conference On Computer Applications Of the Seventh International Conference On Computer ApplicationsYangon, Myanmar 2009 * VDhanalakshmi AnandKumar MRekha RU ArunKumar CSoman KPRajendran S Morphological Analyzer for Agglutinative Languages * Hidden Markov Model with Rule Based Approach for Part of Speech Tagging of Myanmar Language KKZin NLThein 2009 Yangon * Hybride Neuro and Rule-Based Part of Speech Taggers QMa MMurata KUchimoto HIsahar International Conference on Computation Linguistics 2000 * Detecting Inflection Patterns in Natural Language by Minimization of Morphological Model AGelbukh MAlexandrov SYHan Progress