Question answering system with Hidden Markov Model speech recognition

Abstract

Question answering system is a system that can give an answer from the user. In general, question answering can generate answer to text questions. This paper reports the result of question answering system that can receive input questions from speech and text. Hidden Markov Model (HMM) used to recognize the voice provided by the user. The HMM speech recognition used the feature value obtained from Mel Frequency Cepstrum Coefficients method (MFCC). The question answering system used Vector Space Model from Lucene search engine to retrieve relevant documents. The result shows that HMM speech recognition system's success rate in recognizing words is 83.31% which obtained from 13 tested questions. The result also shows that question answering system can answer 4 out of 6 questions that correctly identified by speech recognition system.