Speech Summarization Using Extractive Text Summarization Approach

Abstract

This study describes how extractive text summarising algorithms may be used to accomplish speech-to-text summarization. Our goal is to determine which of the six summarization approach studied in this research is best suited for the job of audio summarization and to provide a suggestion. First, six text summarising methods have been selected: Luhn, LexRank, TextRank, KLSum, LSA, and SumBasic. Then, we analysed them using ROUGE measures on two datasets, DUC2001 and OWIDSum. Then, we picked five voice files from the ISCI Corpus collection and converted them employing the Automatic Speech Recognition (ASR) from the Google API. Findings revealed that Luhn and TextRank performed better at extracting audio summary on the analysed data.