A Decoding Algorithm for Length-Control Summarization Based on Directed Acyclic Transformers

TL;DR

This paper introduces a novel length-control decoding algorithm using Directed Acyclic Transformers and SeqMAP, achieving state-of-the-art results in length-constrained summarization tasks.

Contribution

It proposes a new decoding method that predicts multiple sequence fragments and connects them via a path, improving length control in summarization.

Findings

Achieves state-of-the-art performance on Gigaword and DUC2004 datasets.

Effectively satisfies length constraints with the SeqMAP decoding algorithm.

Utilizes beam search and reranking for performance enhancement.

Abstract

Length-control summarization aims to condense long texts into a short one within a certain length limit. Previous approaches often use autoregressive (AR) models and treat the length requirement as a soft constraint, which may not always be satisfied. In this study, we propose a novel length-control decoding algorithm based on the Directed Acyclic Transformer (DAT). Our approach allows for multiple plausible sequence fragments and predicts a \emph{path} to connect them. In addition, we propose a Sequence Maximum a Posteriori (SeqMAP) decoding algorithm that marginalizes different possible paths and finds the most probable summary satisfying the length budget. Our algorithm is based on beam search, which further facilitates a reranker for performance improvement. Experimental results on the Gigaword and DUC2004 datasets demonstrate our state-of-the-art performance for length-control summarization.