Graph-based Summarization Research Articles

EDCOSUM: Text extractive summarization framework based on edge information with coreference resolution

Automatic Text Summarization(ATS) is distinctly beneficial due to a vast amount of textual data and time-consuming manual summarization. In order to enhance ATS for single document in huge datasets, a new extractive graph framework - text extractive SUMmarization framework based on EDge information with COreference resolution EDCOSUM is proposed in this paper that relies on coreference resolution, adding edge information in word-level graph and a sentence-ranking strategy. EDCOSUM combines the graph-based and statistical-based extractive summarization methods. It is a general method for any document (not limited to a specific domain). Moreover, two ranking strategies(sentence and LSA ranking strategy) are proposed for sentence selection. A set of extensive experiments on CNN/Daily Mail and NEWSROOM are conducted for investigating the proposed method. The widely used automatic evaluation tool: Recall-Oriented Understudy for Gisting Evaluation(ROUGE) is utilized to evaluate EDCOSUM. Compared to the state-of-the-art ATS methods, EDCOSUM achieves a competitive result by improvements of over the highest scores in the literature for metrics ROUGE-1, ROUGE-2 and ROUGE-L respectively.

Enhancing Scientific Papers Summarization with Citation Graph

Previous work for text summarization in scientific domain mainly focused on the content of the input document, but seldom considering its citation network. However, scientific papers are full of uncommon domain-specific terms, making it almost impossible for the model to understand its true meaning without the help of the relevant research community. In this paper, we redefine the task of scientific papers summarization by utilizing their citation graph and propose a citation graph-based summarization model CGSum which can incorporate the information of both the source paper and its references. In addition, we construct a novel scientific papers summarization dataset Semantic Scholar Network (SSN) which contains 141K research papers in different domains and 661K citation relationships. The entire dataset constitutes a large connected citation graph. Extensive experiments show that our model can achieve competitive performance when compared with the pretrained models even with a simple architecture. The results also indicates the citation graph is crucial to better understand the content of papers and generate high-quality summaries.

A new graph-based extractive text summarization using keywords or topic modeling

In graph-based extractive text summarization techniques, the weight assigned to the edges of the graph is the crucial parameter for the sentence ranking. The weights associated with the edges are based on the similarity between sentences (nodes). Most of the graph-based techniques use the common words based similarity measure to assign the weight. In this paper, we propose a new graph-based summarization technique, which, besides taking into account the similarity among the individual sentences, also considers the similarity between the sentences and the overall (input) document. While assigning the weight among the edges of the graph, we consider two attributes. The first attribute is the similarity among the nodes, which forms the edges of the graph. The second attribute is the weight given to a component that represents how much the particular edge is similar to the topics of the overall document for which we incorporate the topic modeling. Along with these modifications, we use the semantic measure to find the similarity among the nodes. The evaluation results of the proposed method demonstrate a significant improvement of the summary quality over the existing text summarization techniques.

Summarization of biomedical articles using domain-specific word embeddings and graph ranking

Text summarization tools can help biomedical researchers and clinicians reduce the time and effort needed for acquiring important information from numerous documents. It has been shown that the input text can be modeled as a graph, and important sentences can be selected by identifying central nodes within the graph. However, the effective representation of documents, quantifying the relatedness of sentences, and selecting the most informative sentences are main challenges that need to be addressed in graph-based summarization. In this paper, we address these challenges in the context of biomedical text summarization. We evaluate the efficacy of a graph-based summarizer using different types of context-free and contextualized embeddings. The word representations are produced by pre-training neural language models on large corpora of biomedical texts. The summarizer models the input text as a graph in which the strength of relations between sentences is measured using the domain specific vector representations. We also assess the usefulness of different graph ranking techniques in the sentence selection step of our summarization method. Using the common Recall-Oriented Understudy for Gisting Evaluation (ROUGE) metrics, we evaluate the performance of our summarizer against various comparison methods. The results show that when the summarizer utilizes proper combinations of context-free and contextualized embeddings, along with an effective ranking method, it can outperform the other methods. We demonstrate that the best settings of our graph-based summarizer can efficiently improve the informative content of summaries and decrease the redundancy.

Graph-based biomedical text summarization: An itemset mining and sentence clustering approach

ObjectiveAutomatic text summarization offers an efficient solution to access the ever-growing amounts of both scientific and clinical literature in the biomedical domain by summarizing the source documents while maintaining their most informative contents. In this paper, we propose a novel graph-based summarization method that takes advantage of the domain-specific knowledge and a well-established data mining technique called frequent itemset mining. MethodsOur summarizer exploits the Unified Medical Language System (UMLS) to construct a concept-based model of the source document and mapping the document to the concepts. Then, it discovers frequent itemsets to take the correlations among multiple concepts into account. The method uses these correlations to propose a similarity function based on which a represented graph is constructed. The summarizer then employs a minimum spanning tree based clustering algorithm to discover various subthemes of the document. Eventually, it generates the final summary by selecting the most informative and relative sentences from all subthemes within the text. ResultsWe perform an automatic evaluation over a large number of summaries using the Recall-Oriented Understudy for Gisting Evaluation (ROUGE) metrics. The results demonstrate that the proposed summarization system outperforms various baselines and benchmark approaches. ConclusionThe carried out research suggests that the incorporation of domain-specific knowledge and frequent itemset mining equips the summarization system in a better way to address the informativeness measurement of the sentences. Moreover, clustering the graph nodes (sentences) can enable the summarizer to target different main subthemes of a source document efficiently. The evaluation results show that the proposed approach can significantly improve the performance of the summarization systems in the biomedical domain.

Text Summarization of Multi-Aspect Comments in Social Networks in Persian Language

Now-a-days, there are increasingly huge amount of user generated comments on the web. The user generated comments usually contains useful and essential information reflecting public’s or customers’ opinions. Since the information in the comments could be used for decision making, production or service improvement, and achieving user satisfaction, the systematic analysis of these comments is an essential need in so many domains including e-commerce, production, and social network analysis. However, the analysis of large volume of comments is a difficult and time-consuming task. Therefore, the need for a system which can convert this massive volume of comments to a useful and efficient summary is felt more and more. Text summarization leads to using more resources at higher speeds and getting richer information. According to numerous studies conducted in the field of multi-document summarization, few studies can be found that have been focused on the user generated comments in Persian language. In this paper, we propose a novel approach to summarize huge amount of comments in Persian, which is enough close to a human summarization. Our approach is based on semantic and lexical similarities and uses a graph-based summarization. We also propose a clustering to deal with multiple aspects (subjects) in a corpus of comments. According to the experiments, the summaries extracted by the proposed approach reached an average score of 8.75 out of 10, which improves the state-of-the-art summarizer’s score about 14 percent.

Search engine reinforced semi-supervised classification and graph-based summarization of microblogs

There is an abundance of information found on microblog services due to their popularity. However the potential of this trove of information is limited by the lack of effective means for users to browse and interpret the numerous messages found on these services. We tackle this problem using a two-step process, first by slicing up the search results of current retrieval systems along multiple possible genres. Then, a summary is generated from the microblog messages attributed to each genre. We believe that this helps users to better understand the possible interpretations of the retrieved results and aid them in finding the information that they need. Our novel approach makes use of automatically acquired information from external search engines in each of these two steps. We first integrate this information with a semi-supervised probabilistic graphical model, and show that this helps us to achieve significantly better classification performance without the need for much training data. Next we incorporate the extra information into graph-based summarization, and demonstrate that superior summaries (up to 30% improvement in ROUGE-1) are obtained.

GraphSum: Discovering correlations among multiple terms for graph-based summarization

Graph-based summarization entails extracting a worthwhile subset of sentences from a collection of textual documents by using a graph-based model to represent the correlations between pairs of document terms. However, since the high-order correlations among multiple terms are disregarded during graph evaluation, the summarization performance could be limited unless integrating ad hoc language-dependent or semantics-based analysis.This paper presents a novel and general-purpose graph-based summarizer, namely GraphSum (Graph-based Summarizer). It discovers and exploits association rules to represent the correlations among multiple terms that have been neglected by previous approaches. The graph nodes, which represent combinations of two or more terms, are first ranked by means of a PageRank strategy that discriminates between positive and negative term correlations. Then, the produced node ranking is used to drive the sentence selection process.The experiments performed on benchmark and real-life documents demonstrate the effectiveness of the proposed approach compared to many state-of-the-art summarizers.

Clustering cliques for graph-based summarization of the biomedical research literature.

BackgroundGraph-based notions are increasingly used in biomedical data mining and knowledge discovery tasks. In this paper, we present a clique-clustering method to automatically summarize graphs of semantic predications produced from PubMed citations (titles and abstracts).ResultsSemRep is used to extract semantic predications from the citations returned by a PubMed search. Cliques were identified from frequently occurring predications with highly connected arguments filtered by degree centrality. Themes contained in the summary were identified with a hierarchical clustering algorithm based on common arguments shared among cliques. The validity of the clusters in the summaries produced was compared to the Silhouette-generated baseline for cohesion, separation and overall validity. The theme labels were also compared to a reference standard produced with major MeSH headings.ConclusionsFor 11 topics in the testing data set, the overall validity of clusters from the system summary was 10% better than the baseline (43% versus 33%). While compared to the reference standard from MeSH headings, the results for recall, precision and F-score were 0.64, 0.65, and 0.65 respectively.

TSGVi: a graph-based summarization system for Vietnamese documents

This paper proposes an automatic method to generate an extractive summary of multiple Vietnamese documents which are related to a common topic by modeling text documents as weighted undirected graphs. It initially builds undirected graphs with vertices representing the sentences of documents and edges indicate the similarity between sentences. Then, by adopting PageRank algorithm, we can generate salient scores for sentences. Sentences are ranked according to their salient scores and selected based on maximal marginal relevance to form the summaries. These summaries are combined and applied the same process one more time to form the final extractive summary of the document set. A series of experiments are performed on Vietnamese news articles and English data of DUC 2002, 2003, 2007. The results demonstrate the effectiveness of the proposed technique over reference systems.