PREDICTING SIMILARITY IN SENTENCES THROUGH WORD EMBEDDING

Abstract

In the field of natural language processing, learning the context from a given sentence is a very important and challenging task. Which is great source for predicting the intention of user, this prediction will help to detect the fake NEWS, for creating more interactive artificial intelligent bot that will interact better respond better act better, for giving the better recommendations such as recommending music for that purpose word embedding in used to bridge the gap between the computing machine and the real world. In this research we have examined the various existing models i.e, regression models like multilinear regression, support vector machine, random forest, match LSTM to detect the similar sentences. Also, we compared their results based on accuracy achieved. Moreover, we proposed new model based on convolution neural network warping with time distributed layer which outperform with respect to other models from 77.67% to 83.72%