Large-Margin Learning of Submodular Summarization Methods

TL;DR

This paper introduces a supervised large-margin learning approach for submodular functions in extractive multi-document summarization, improving performance over manually tuned methods by directly optimizing a convex relaxation of the performance measure.

Contribution

It presents a novel large-margin training method applicable to all submodular summarization functions, enabling automatic learning of high-fidelity models with many parameters.

Findings

Significant performance improvements over state-of-the-art manually tuned functions

Effective for both pairwise and coverage-based scoring functions

Applicable across multiple datasets

Abstract

In this paper, we present a supervised learning approach to training submodular scoring functions for extractive multi-document summarization. By taking a structured predicition approach, we provide a large-margin method that directly optimizes a convex relaxation of the desired performance measure. The learning method applies to all submodular summarization methods, and we demonstrate its effectiveness for both pairwise as well as coverage-based scoring functions on multiple datasets. Compared to state-of-the-art functions that were tuned manually, our method significantly improves performance and enables high-fidelity models with numbers of parameters well beyond what could reasonbly be tuned by hand.