A Model For Recapitulating Audio Messages Using Machine Learning

This model aims to develop an efficient way to recapitulate large audio messages or clips for valuable insights. With increase in utilization of audio/visual data day by day, there is a need to handle audio files more intelligently. In this document, a novel approach is presented to build a summarized audio for a given long audio file. This method is composed primarily of three modules namely: Conversion of Speech into Text, Text summarization, and lastly conversion of text into speech. Each module is fed by the output of another module except speech to text conversion where input is the given audio file for which summary has to be formed. The first step in audio recapitulation is conversion of given audio to text. This is made possible by sending asynchronous requests to Google Cloud speech API. The next module accomplishes its task of extracting important sentences from the transcript by using the Text Rank algorithm. The last module is to convert the summarized text generated from the output of text summarization module to an audio file. This whole method is given a suitable User Interface using flask and thus a web application is formed for helping users to interact with this model.