An Adaptive Algorithm for Polysemous Words in Natural Language Processing

Abstract

In this research article the supervised adaptive approach for word sense disambiguation is discussed. The most critical and identified problem of natural language understanding is the lexical ambiguity. Lexical ambiguity is introduced by the polysemy words. This paper describes different state-of-the-art techniques to detect ambiguous words from the ambiguous sentence. The word embedding is an important phase of word sense disambiguation and which is succeeded by ambiguous word detection and processed by classification. The different embedding techniques are discussed here and the uniqueness of adaptive word embedding has been proved. There are two standard available datasets OMSTI and WordNet are used for the data processing. This article elaborates the newly generated dataset Adaptive-Lex for disambiguation. There are two important challenges in word sense disambiguation: identifying polysemy words without context information and constructing word embedding for polysemy words with highest sense values. These challenges are addressed with a complex network approach and adaptive word embedding technique. The performance evaluation classifiers are supervised neural network approaches such CNN and DNN.