Abstract:Extractive document summarization is a fundamental task in natural language processing (NLP). Recently, several Graph Neural Networks (GNNs) are proposed for this task. However, most existing GNN-based models can neither effectively encode semantic nodes of multiple granularity level apart from sentences nor substantially capture different cross-sentence meta-paths. To address these issues, we propose MHgatSum, a novel Multi-granularity Heterogeneous Graph ATtention networks for extractive document SUMmarization. Specifically, we first build a multi-granularity heterogeneous graph (HetG) for each document, which is better to represent the semantic meaning of the document. The HetG contains not only sentence nodes but also multiple other granularity effective semantic units with different semantic levels, including keyphrases and topics. These additional nodes act as the intermediary between sentences to build the meta-paths involved in sentence node (i.e., Sentence-Keyphrase-Sentence and Sentence-Topic-Sentence). Then, we propose a heterogeneous graph attention networks to embed the constructed HetG for extractive summarization, which enjoys multi-granularity semantic representations. The model is based on a hierarchical attention mechanism, including node-level and semantic-level attentions. The node-level attention can learn the importance between a node and its meta-path based neighbors, while the semantic-level attention is able to learn the importance of different meta-paths. Moreover, to better integrate sentence global knowledge, we further incorporate sentence node global importance in local node-level attention. We conduct empirical experiments on two benchmark datasets, which demonstrates the superiority of MHgatSum over previous SOTA models on the task of extractive summarization.

Multi-layered graph-based multi-document summarization model

Query-focused Multi-document Summarization: Combining a Novel Topic Model with Graph-based Semi-supervised Learning

Exploring hypergraph-based semi-supervised ranking for query-oriented summarization

A Document-Sensitive Graph Model for Multi-Document Summarization

A Multi-Granularity Heterogeneous Graph for Extractive Text Summarization

Multi-Roles Graph Based Extractive Summarization

An Integrated Graph Model For Document Summarization

Graph-based Neural Multi-Document Summarization

Leveraging Graph to Improve Abstractive Multi-Document Summarization.

Multi-document Summarization Via Sentence-Level Semantic Analysis and Symmetric Matrix Factorization

Query-focused Multi-Document Summarization: Combining a Topic Model with Graph-based Semi-supervised Learning.

Multi-granularity heterogeneous graph attention networks for extractive document summarization

Multi-Document Summarization Based On Two-Level Sparse Representation Model

Multi-Document Abstractive Summarization Using Chunk-graph and Recurrent Neural Network

Research on Automatic Multi--document Summarization Based on Concept CO-occurrence Graph

Multiplex Graph Neural Network for Extractive Text Summarization

SgSum: Transforming Multi-document Summarization into Sub-graph Selection

Parallel Relationship Graph to Improve Multi-Document Summarization

An Exploration of Document Impact on Graph-Based Multi-Document Summarization

SentTopic-MultiRank: a Novel Ranking Model for Multi-Document Summarization.

A Study of Unsupervised Multi-document Sentiment Summarization Techniques Based on Graph Structure