A Bangla Word Sense Disambiguation Technique using Minimum Edit Distance Algorithm and Cosine Distance
Protap Kumar Saha, Anamika Das Mou, Tanni Mittra · 2019
In Natural Language Processing, Morphology known as the most decisive part. It can be more difficult when there are several meanings for only one word. Ambiguous word is a word which has those several meanings. The human brain can easily identify these ambiguities but for machines, it is very complicated to detect. Word Sense Disambiguation(WSD) is such a technique that trains machines to detect ambiguities. Different types of research work have been published in different languages for this technique. But developing an optimized Bangla WSD system is still a great research challenge. To overcome this challenge we have to propose a new technique to detect ambiguous word in a sentence. A corpus containing 3860 sentences is built from different resources. We applied the Levenshtein distance algorithm to detect ambiguous word and Cosine Similarity to sense actual meaning in a given Bangla sentence. The accuracy of our method is 80.82%. The validity of our claims has been proved through comparisons with other well established methods.