Zeitschrift für Informatik und Systembiologie

Comprehensive Analysis of Word Sensing Tool and Techniques for Enhancing Classification Accuracy of Query String

Abstract

Sunita Mahajan and Vijay Rana

Word sense in the field of natural language processing (NLP) is a corner stone for appropriate word selection. A word can contain more than one sense, but machine can’t extract the actual sense of the given particular content. Implication of this situation is mismatch between the user requirements and result generates through the machine. e.g., User wants to search a query “What is word sensing?” The machine can’t find the relation between these two words “word”, “sensing”. Relationship between words cannot be extracted by the machine and more results corresponding to sensing is displayed and user requirements corresponding to “Word Sensing” as a whole are rejected. primary reason for this mismatch is due to static dictionary possessed by web servers. Techniques we are analysis different types of techniques and algorithms for the word sense. The major techniques are which used to word sense are knowledge based approaches are based on different knowledge sources as machine readable dictionaries to extract the sense like thesauri, Word net are machine readable dictionaries to find the word sense, Supervised learning technique is a manually extract the sense from the data. In this process trained the target words through the labelling, unsupervised learning technique in this process words are no needs to be trained target data are based on the clustering, Semi-Supervised learning technique is a hybrid approach of the supervised and unsupervised. In this process target words are based on the particular content. Tools for building word database to be accessed by the web applications including Word Net, Image Net and Babel Net are discussed in this literature. Our Contribution we conduct comprehensive review of knowledge based, supervised, unsupervised and semisupervised learning techniques used in the field of word sensing and detect the best word sensing mechanism for fetching only relevant material from the web while decreasing the execution time for content retrieval.