Reflexive hybrid approach to provide precise answer of user desired frequently asked question

To answer consumers' queries as per their expectation has become a need and requirement in today's scenario. On the other end, it is a challenging problem to handle and acquire precise answer of everyone's query. In several commercial systems question-answer section has become a reflexive and robotic system which handles answers associated to users questions with a newfangled idea. In today's scenario frequent asked question section are not static questions and answer as in earlier system, therefore system has to answer every query by matching the most similar question from the archives or provide relevant content from website. Same issue has been attended here and proposed a novel approach in order to improve its ability and overcome the challenging task of finding similar questions from the knowledge base of files of publicly available frequently asked questions accurately. In the proposed approach, we extract lexical, structural and semantic behavior of the user query and provide a hybrid approach which is basically combination of term frequency-Inverse document frequency with POS tagging and Word2Vec to retrieve most similar answer corresponding to entered user query. These techniques include methods taken from the natural language processing and information retrieval. We have also compared our approaches using cosine similarity as an evaluation metric along with considering the time complexity. Experimental results shows that our proposed hybrid approach gives better performance as compared to other individual techniques.