Unsupervised Extractive Summarization with Heterogeneous Graph Embeddings for Chinese Document

TL;DR

This paper introduces an unsupervised extractive summarization method for Chinese documents using heterogeneous graph embeddings, capturing rich word-sentence interactions and outperforming existing baselines.

Contribution

It is the first to utilize heterogeneous graph embeddings for unsupervised Chinese extractive summarization, integrating diverse interaction types and additional nodes like keywords.

Findings

Outperforms strong baselines on three datasets

Effectively captures complex word-sentence interactions

Flexible graph structure allows easy integration of new node types

Abstract

In the scenario of unsupervised extractive summarization, learning high-quality sentence representations is essential to select salient sentences from the input document. Previous studies focus more on employing statistical approaches or pre-trained language models (PLMs) to extract sentence embeddings, while ignoring the rich information inherent in the heterogeneous types of interaction between words and sentences. In this paper, we are the first to propose an unsupervised extractive summarizaiton method with heterogeneous graph embeddings (HGEs) for Chinese document. A heterogeneous text graph is constructed to capture different granularities of interactions by incorporating graph structural information. Moreover, our proposed graph is general and flexible where additional nodes such as keywords can be easily integrated. Experimental results demonstrate that our method consistently outperforms the strong baseline in three summarization datasets.