SentBS: Sentence-level Beam Search for Controllable Summarization

TL;DR

SentBS introduces a sentence-level beam search method that improves control over the structure of generated summaries, significantly reducing structural discrepancies compared to existing models.

Contribution

The paper proposes SentBS, a novel sentence-level beam search approach that enhances structure control in summarization, addressing limitations of previous methods.

Findings

All SentBS combinations improved structure agreement.

The best method reduced structural discrepancies by approximately 68%.

SentBS outperforms existing structure-controlled summarization models.

Abstract

A wide range of control perspectives have been explored in controllable text generation. Structure-controlled summarization is recently proposed as a useful and interesting research direction. However, current structure-controlling methods have limited effectiveness in enforcing the desired structure. To address this limitation, we propose a sentence-level beam search generation method (SentBS), where evaluation is conducted throughout the generation process to select suitable sentences for subsequent generations. We experiment with different combinations of decoding methods to be used as subcomponents by SentBS and evaluate results on the structure-controlled dataset MReD. Experiments show that all explored combinations for SentBS can improve the agreement between the generated text and the desired structure, with the best method significantly reducing the structural discrepancies suffered by the existing model, by approximately 68%.