Intelligent meeting recording system and operating method

Abstract

An intelligent meeting recording system and an operating method are provided. The intelligent meeting recording system firstly receives an audio file and a meeting agenda, and then converts the audio file to a text by a speech-to-text model so as to generate a transcription. The transcription can be trimmed by a language model, and speakers in the transcription can be separated. After key information of the meeting is obtained according to the agenda, the language model relies on the key information to process the transcription and a bigram model is used to calculate importance for each of words in multiple key paragraphs. The language model then relies on the importance of each of the words in the key paragraphs to generate a meeting summary. The information of time and people of the transcription, matters to be assigned and the key information in the agenda are referred to for establishing a to-do list.

Images