基于查询的可控会议摘要算法研究

Meeting summaries can efficiently convey the main content of the meeting to participants/non-participants. However, a meeting often involves multiple topics, and different roles have different divisions of labor, they usually need to focus on different parts of the meeting. Therefore, only creating a general meeting summary cannot well meet the needs of different users. Aiming at the above problems, this paper proposes a query-based meeting summarization model. Firstly, We use a correlation model to calculate the relevance score of each round of turn in the meeting to the input query, and then the relevance score is considered in the meeting summarization model. On the basis of existing research, this paper innovatively proposes a mechanism combining query relevance and summarization models. Experiments show that our model achieves competitive performance on the QMSum dataset.