Word Sense Disambiguation for Assamese

Word Sense Disambiguation (WSD) is the process of identifying the proper sense of an ambiguous word depending on the particular context. It is to find the accurate sense si among the set of senses {s₁, s₂,..., s_n}. This task was motivated by its interpretation in various Natural Language Processing (NLP) applications like IR, MT, QA, TC, SP etc. In this paper, machine learning technique - Naive Bayes Classifier was used for automatic disambiguation task. Training data was prepared with sense annotated features. For preparing sense annotated data we took help of the sense inventory. Currently, about 160 ambiguous words are present in the sense inventory derived from 18K and 25K words from Assamese Corpus and WordNet. The system is implemented in two phases. In the first phase, a total of 2.7K sense annotated training data and 800 test data were taken and a result of 71% accuracy was found. Analyzing the result depicts that accuracy improves as the training data size gradually increases and by the learned model generated in the previous iteration. In second phase we manually validate the outcomes of first-phase and we add those clean sense tagged data to previous training data set. Than we train our system with our increasing training data (3.5K) which enhance the result accuracy. An iterative learning is adopted by the system and more accuracy of 7% is achieved. This paper aims to implement Assamese WSD system by NB classifier using lexical features and enhancement of the baseline method turns out in improving the classifier accuracy to 78%.