A Focused Study on Sequence Length for Dialogue Summarization
This work addresses the specific issue of summary length control for dialogue summarization systems, which is incremental in nature.
The study tackled the problem of output length in dialogue summarization by analyzing model verbosity, identifying features for length prediction, and showing that a length-aware summarizer improves existing models, achieving notable improvements on DialogSum and SAMSum datasets.
Output length is critical to dialogue summarization systems. The dialogue summary length is determined by multiple factors, including dialogue complexity, summary objective, and personal preferences. In this work, we approach dialogue summary length from three perspectives. First, we analyze the length differences between existing models' outputs and the corresponding human references and find that summarization models tend to produce more verbose summaries due to their pretraining objectives. Second, we identify salient features for summary length prediction by comparing different model settings. Third, we experiment with a length-aware summarizer and show notable improvement on existing models if summary length can be well incorporated. Analysis and experiments are conducted on popular DialogSum and SAMSum datasets to validate our findings.