StreamHover: Livestream Transcript Summarization and Annotation

TL;DR

StreamHover introduces a large annotated dataset and a neural extractive summarization model tailored for livestream transcripts, addressing challenges posed by informal spoken language and enabling efficient content browsing.

Contribution

The paper presents a new annotated dataset of over 500 hours of livestream transcripts and a novel neural extractive summarization model using vector-quantized variational autoencoders.

Findings

Model outperforms strong baselines in summarization tasks.

Large-scale annotated dataset enhances research in livestream transcript summarization.

Proposes a generalizable approach for summarizing informal spoken content.

Abstract

With the explosive growth of livestream broadcasting, there is an urgent need for new summarization technology that enables us to create a preview of streamed content and tap into this wealth of knowledge. However, the problem is nontrivial due to the informal nature of spoken language. Further, there has been a shortage of annotated datasets that are necessary for transcript summarization. In this paper, we present StreamHover, a framework for annotating and summarizing livestream transcripts. With a total of over 500 hours of videos annotated with both extractive and abstractive summaries, our benchmark dataset is significantly larger than currently existing annotated corpora. We explore a neural extractive summarization model that leverages vector-quantized variational autoencoder to learn latent vector representations of spoken utterances and identify salient utterances from the transcripts to form summaries. We show that our model generalizes better and improves performance over strong baselines. The results of this study provide an avenue for future research to improve summarization solutions for efficient browsing of livestreams.