Abstractive Meeting Summarization: A Survey

Abstract

Recent advances in deep learning, and especially the invention of encoder-decoder architectures, has significantly improved the performance of abstractive summarization systems. While the majority of research has focused on written documents, we have observed an increasing interest in the summarization of dialogues and multi-party conversation over the past few years. A system that could reliably transform the audio or transcript of a human conversation into an abridged version that homes in on the most important points of the discussion would be valuable in a wide variety of real-world contexts, from business meetings to medical consultations to customer service calls. This paper focuses on abstractive summarization for multi-party meetings, providing a survey of the challenges, datasets and systems relevant to this task and a discussion of promising directions for future study.

Action-Item-Driven Summarization of Long Meeting Transcripts

“…Abstractive summarization techniques focus more on understanding the overall meaning of a transcript and then generating a concise summary based on the entire text. Unlike extractive summarization, abstractive summarization generates new words and phrases that were not found in the input transcript, rather than simply extracting the important phrases [18]. Abstractive summarization is more challenging, but as expected, it leads to better summaries [9].…”

“…However, ROUGE scores have many flaws since they focus solely on the lexical overlap between the machine-generated summaries and the human reference summaries rather than their semantic similarity [5]. As a result, BERTScore, which measures the semantic similarity between the machine-generated summaries and the reference summaries has been growing in popularity [18]. We employ the BERTScore metric as well, since it has been shown to achieve higher correlations with human judgment on the quality of a machine-generated summary compared to ROUGE [27].…”

Learning from Failure: Improving Meeting Summarization without Good Samples

“…Automatic meeting summarization aims to condense information from a large piece of meeting transcript and produce a concise and comprehensible minutes or digest automatically (Gillick et al 2009;Kumar and Kabiri 2022). While earlier works focus on extractive methods that create summaries by directly selecting and concatenating unedited sentences from source text, the development of neural networks has encouraged a growing trend in abstractive methods that implement encoder-decoder architectures on source text to gener-ate summaries (Shang et al 2018;Rennard et al 2023). The lengthy, multi-speaker, spoken-language natures of meeting text pose many challenges for summarization, several strategies are proposed to address different aspects of those challenges (Li et al 2019;Zhu et al 2020;Koay et al 2021;Zou et al 2021).…”

“…The widespread use of video-telephony applications and the surging trend of remote work have contributed to a boom in online chatting and virtual meetings (Rennard et al 2023). As a result, automatic meeting summarization, which condenses long and unorganized meeting transcripts into concise and comprehensible summaries, is of increasingly potential social economic values and thus draws rising attention (Gillick et al 2009;Shang et al 2018).…”