An Exploratory Study on Long Dialogue Summarization: What Works and What's Next

TL;DR

This study explores methods for long dialogue summarization, comparing extended transformers, retrieval-based, and hierarchical models, and finds retrieve-then-summarize approaches perform best on multiple datasets.

Contribution

It provides a comprehensive comparison of three strategies for long dialogue summarization and highlights the effectiveness of retrieve-then-summarize pipelines with improved retrieval and pretraining.

Findings

Retrieve-then-summarize models outperform other strategies.

Stronger retrieval models improve summary quality.

Pretraining on external datasets enhances performance.

Abstract

Dialogue summarization helps readers capture salient information from long conversations in meetings, interviews, and TV series. However, real-world dialogues pose a great challenge to current summarization models, as the dialogue length typically exceeds the input limits imposed by recent transformer-based pre-trained models, and the interactive nature of dialogues makes relevant information more context-dependent and sparsely distributed than news articles. In this work, we perform a comprehensive study on long dialogue summarization by investigating three strategies to deal with the lengthy input problem and locate relevant information: (1) extended transformer models such as Longformer, (2) retrieve-then-summarize pipeline models with several dialogue utterance retrieval methods, and (3) hierarchical dialogue encoding models such as HMNet. Our experimental results on three long dialogue datasets (QMSum, MediaSum, SummScreen) show that the retrieve-then-summarize pipeline models yield the best performance. We also demonstrate that the summary quality can be further improved with a stronger retrieval model and pretraining on proper external summarization datasets.