Reducing Error Rate for ASR using Semantic Error Correction Approach

Abstract

Many application environments have already used speech interface. But the low speech recognition rate makes it difficult to extend its application to new fields. In the human-computer interaction through spoken dialogue are being investigated. Automatic Speech recognition (ASR) is the process of converting a spoken speech into text that can be manipulated by the computer. The state of the art in automatic speech recognition has reached the point that searching for and extracting information from large speech repositories. This system presents semantic-oriented approach to correct both semantic and lexical errors, which is also more accurate for especially domainspecific speech error correction. This paper demonstrates the superior performance of this approach and some advantages over previous lexicaloriented approaches by comparing such approaches. Experiments carried out on various speeches in English syllable indicated a successful decrease in the number of errors and an improvement in overall error correction rate.