Knowledge Extraction and Question Formation from Transcripts of Video using WebRTC and YAKE

Abstract

With the advancements in communication technology we now have access to more sophisticated and effective methods of communication. People can interact with one another through electronic mediums and this has paved the way for corporations and businesses to connect better with partners and employees. Video conferencing is a type of distance conferencing where two or more number of people can visually connect and interact with each other. The face of video communication has changed dramatically due to the global pandemic. The ban on travel has limited the reach of the enterprises in the foreign market, which in turn, has increased the adoption of video conferencing software.If some users miss the conference or get disconnected or want to review the conference again ,going through the recorded video of the conference becomes a hassle. Also there is a problem of language barrier in video conferencing. The presented system overcomes all these problems. The system firstly generates the transcript of the meeting in language which was used in the meeting by using speech to text recognition and stores it. Some meetings may be conducted for a long duration of time hence reading these transcripts becomes time consuming and is not feasible. So to handle these situations the system generates a summary of each meeting by extracting the important sections of the meeting by using YAKE keyword extractor. In addition to this a question answer based format is also generated in the selected language so the users may search the questions they want the answer to ,instead of reading the whole summary which saves time. The presented system helps the end user by understanding the essential part of the meeting in their preferred language.