LexRank: Graph-based Lexical Centrality as Salience in Text Summarization

By Günes Erkan, Dragomir R. Radev. 2004

3,275 citations recorded by the source corpus.

We introduce a stochastic graph-based method for computing relative importance of textual units for Natural Language Processing. We test the technique on the problem of Text Summarization (TS). Extractive TS relies on the concept of sentence salience to identify the most important sentences in a document or set of documents. Salience is typically defined in terms of the presence of particular important words or in terms of similarity to a centroid pseudo-sentence. We consider a new approach, LexRank, for computing sentence importance based on the concept of eigenvector centrality in a graph representation of sentences. In this model, a connectivity matrix based on intra-sentence cosine similarity is used as the adjacency matrix of the graph representation of sentences. Our system, based on LexRank ranked in first place in more than one task in the recent DUC 2004 evaluation. In this paper we present a detailed analysis of our approach and apply it to a larger data set including data from earlier DUC evaluations. We discuss several methods to compute centrality using the similarity graph. The results show that degree-based methods (including LexRank) outperform both centroid-based methods and other systems participating in DUC in most of the cases. Furthermore, the LexRank with threshold method outperforms the other degree-based techniques including continuous LexRank. We also show that our approach is quite insensitive to the noise in the data that may result from an imperfect topical clustering of documents.

Subfields: Artificial Intelligence, Molecular Biology, Natural Language Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Molecular Biology

In Natural Language Processing

Citations in the corpus

Big Bird: Transformers for Longer Sequences (2020, neutral): “66 LexRank [26] 33.”
Learning to Extract Coherent Summary via Deep Reinforcement Learning (2018, methodology): “Traditional methods mainly focus on scoring sentences using graph-based method (Erkan and Radev 2004), submodular functions (Lin and Bilmes 2011) or integer linear programming (Berg-Kirkpatrick, Gillick, and Klein 2011), which are coupled with handcrafted features.”
A Discourse-Aware Attention Model for Abstractive Summarization of Long Documents (2018, methodology): “The compared methods are: LexRank (Erkan and Radev, 2004), SumBa-sic (Vanderwende et al., 2007), LSA (Steinberger and Jezek, 2004), Attn-Seq2Seq (Nallapati et al., 2016; Chopra et al., 2016), Pntr-Gen-Seq2Seq (See et al., 2017).”
Determinantal Point Processes for Machine Learning (2012, methodology): “The node weights are given by the LexRank score of each paper [43].”
Regularizing Text Categorization with Clusters of Words (2016, methodology): “Graph-of-words was originally an idea of Mihalcea and Tarau (2004) and Erkan and Radev (2004) who applied it to the tasks of unsupervised keyword extraction and extractive single document summarization.”
Ranking Sentences for Extractive Summarization with Reinforcement Learning (2018, background): “A vast majority of these methods learn to score each sentence independently (Barzilay and Elhadad, 1997; Teufel and Moens, 1997; Erkan and Radev, 2004; Mihalcea and Tarau, 2004; Shen et al., 2007; Schilder and Kondadadi, 2008; Wan, 2010) and a summary is generated by selecting top-scored sentences i”
SemEval-2016 Task 1: Semantic Textual Similarity, Monolingual and Cross-Lingual Evaluation (2016, methodology): “The team extends a multi-perspective convolutional neural network (MPCNN) (He et al., 2015) with a simple word level attentional mecha- 17To see how much of this is related to the Q&A domain in particular, we will investigate including difficult non-Q&A evaluation data in future STS…”
BIGPATENT: A Large-Scale Dataset for Abstractive and Coherent Summarization (2019, methodology): “For L EX R ANK and S UM B ASIC , we used sumy 8 .”
This Email Could Save Your Life: Introducing the Task of Email Subject Line Generation (2019, background): “Other systems can achieve higher ME-TEOR scores because METEOR emphasizes re-call (recall weighted 9 times more than precision) and other extractive systems such as LexRank can generate longer sentences as subject lines.”
Neural Document Summarization by Jointly Learning to Score and Select Sentences (2018, methodology): “Graph-based methods such as TextRank (Mihal-cea and Tarau, 2004) and LexRank (Erkan and Radev, 2004) measure sentence importance using weighted-graphs.”
On Extractive and Abstractive Neural Document Summarization with Transformer Language Models (2020, methodology): “These methods include LexRank (Erkan and Radev, 2004), SumBasic (Van-derwende et al., 2007), LSA (Steinberger and Jezek, 2004), Attention-Seq2Seq (Nallapati et al., 2016a; Chopra et al., 2016), Pointer-Generator Seq2Seq (See et al., 2017), Discourse-aware , which is a hierarchical extension to the…”
Improving Zero and Few-Shot Abstractive Summarization with Intermediate Fine-tuning and Data Augmentation (2020, methodology): “work, extractive models such as TextRank, (Mihalcea and Tarau, 2004), LexRank (Erkan and Radev, 2004), and more recently PacSum (Zheng and Lapata, 2019), aim to use graph centrality in order to extract important sentences for a document.”
Abstractive Document Summarization with a Graph-Based Attentional Neural Model (2017, methodology): “To tackle this challenge, we learn from graph-based extractive summarization models TextRank (Mihalcea and Tarau, 2004) and LexRank (Erkan and Radev, 2004), which are based on the PageR-ank (Page et al., 1999) algorithm.”
Heterogeneous Graph Neural Networks for Extractive Document Summarization (2020, methodology): “Early works focus on document graphs constructed with the content similarity among sentences, like LexRank (Erkan and Radev, 2004) and TextRank (Mihalcea and Tarau, 2004).”
Variations of the Similarity Function of TextRank for Automated Summarization (2016, methodology): “The Tex-tRank algorithm developed by Mihalcea and Tarau [17] and the LexRank al-gorithm by Erkan and Radev [7] are based in ranking the lexical units of the text (sentences or words) using variations of PageRank [20].”
Sentence similarity based on semantic nets and corpus statistics (2006, background): “In addition, the incorporation of short-text similarity is beneficial to applications such as text summarization [9], text categorization [15], and machine translation [21].”