CUED_speech at TREC 2020 Podcast Summarisation Track
This work provides a strong solution for automatically summarizing podcasts, which is beneficial for listeners and content platforms seeking efficient content consumption.
This paper addresses the problem of podcast summarization by filtering redundant sentences using a hierarchical attention model and then applying a fine-tuned BART model. Their best system achieved an EGFB score of 1.777, outperforming creator-provided descriptions (1.291) and winning the TREC 2020 Podcast Summarisation Challenge.
In this paper, we describe our approach for the Podcast Summarisation challenge in TREC 2020. Given a podcast episode with its transcription, the goal is to generate a summary that captures the most important information in the content. Our approach consists of two steps: (1) Filtering redundant or less informative sentences in the transcription using the attention of a hierarchical model; (2) Applying a state-of-the-art text summarisation system (BART) fine-tuned on the Podcast data using a sequence-level reward function. Furthermore, we perform ensembles of three and nine models for our submission runs. We also fine-tune the BART model on the Podcast data as our baseline. The human evaluation by NIST shows that our best submission achieves 1.777 in the EGFB scale, while the score of creator-provided description is 1.291. Our system won the Spotify Podcast Summarisation Challenge in the TREC2020 Podcast Track in both human and automatic evaluation.