This paper details the features and the methodology adopted in the construction of the CNN-corpus, a test corpus for single document extractive text summarization of news articles. The current version of the CNN-corpus encompasses 3,000 texts in English, and each of them has an abstractive and an extractive summary. The corpus allows quantitative and qualitative assessments of extractive summarization strategies.
more »
« less
Summarization assessment methodology for multiple corpora using queries and classification for functional evaluation
Extractive summarization is an important natural language processing approach used for document compression, improved reading comprehension, key phrase extraction, indexing, query set generation, and other analytics approaches. Extractive summarization has specific advantages over abstractive summarization in that it preserves style, specific text elements, and compound phrases that might be more directly associated with the text. In this article, the relative effectiveness of extractive summarization is considered on two widely different corpora: (1) a set of works of fiction (100 total, mainly novels) available from Project Gutenberg, and (2) a large set of news articles (3000) for which a ground truthed summarization (gold standard) is provided by the authors of the news articles. Both sets were evaluated using 5 different Python Sumy algorithms and compared to randomly-generated summarizations quantitatively. Two functional approaches to assessing the efficacy of summarization using a query set on both the original documents and their summaries, and using document classification on a 12-class set to compare among different summarization approaches, are introduced. The results, unsurprisingly, show considerable differences consistent with the different nature of these two data sets. The LSA and Luhn summarization approaches were most effective on the database of fiction, while all five summarization approaches were similarly effective on the database of articles. Overall, the Luhn approach was deemed the most generally relevant among those tested.
more »
« less
- Award ID(s):
- 1842577
- NSF-PAR ID:
- 10354049
- Date Published:
- Journal Name:
- Integrated Computer-Aided Engineering
- Volume:
- 29
- Issue:
- 3
- ISSN:
- 1069-2509
- Page Range / eLocation ID:
- 227 to 239
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
The DocEng’19 Competition on Extractive Text Summarization assessed the performance of two new and fourteen previously published extractive text sumarization methods. The competitors were evaluated using the CNN-Corpus, the largest test set available today for single document extractive summarization.more » « less
-
Abstractive summarization models typically learn to capture the salient information from scratch implicitly.Recent literature adds extractive summaries as guidance for abstractive summarization models to provide hints of salient content and achieves better performance.However, extractive summaries as guidance could be over strict, leading to information loss or noisy signals.Furthermore, it cannot easily adapt to documents with various abstractiveness.As the number and allocation of salience content pieces varies, it is hard to find a fixed threshold deciding which content should be included in the guidance.In this paper, we propose a novel summarization approach with a flexible and reliable salience guidance, namely SEASON (SaliencE Allocation as Guidance for Abstractive SummarizatiON).SEASON utilizes the allocation of salience expectation to guide abstractive summarization and adapts well to articles in different abstractiveness.Automatic and human evaluations on two benchmark datasets show that the proposed method is effective and reliable.Empirical results on more than one million news articles demonstrate a natural fifteen-fifty salience split for news article sentences, providing a useful insight for composing news articles.more » « less
-
null (Ed.)Query Biased Summarization (QBS) aims to produce a summary of the documents retrieved against a query to reduce the human effort for inspecting the full-text content of a document. Typical summarization approaches extract a document text snippet that has term overlap with the query and show that to a searcher. While snippets show relevant information in a document, to the best of our knowledge, there does not exist a summarization system that shows what relevant concepts is missing in a document. Our study focuses on the reduction of user effort in finding relevant documents by exposing them to omitted relevant information. To this end, we use a classical approach, DSPApprox, to find terms or phrases relevant to a query. Then we identify which terms or phrases are missing in a document, present them in a search interface, and ask crowd workers to judge document relevance based on snippets and missing information. Experimental results show both benefits and limitations of this approach.more » « less
-
Extractive text summarization aims at extract- ing the most representative sentences from a given document as its summary. To extract a good summary from a long text document, sen- tence embedding plays an important role. Re- cent studies have leveraged graph neural net- works to capture the inter-sentential relation- ship (e.g., the discourse graph) to learn con- textual sentence embedding. However, those approaches neither consider multiple types of inter-sentential relationships (e.g., semantic similarity & natural connection), nor model intra-sentential relationships (e.g, semantic & syntactic relationship among words). To ad- dress these problems, we propose a novel Mul- tiplex Graph Convolutional Network (Multi- GCN) to jointly model different types of rela- tionships among sentences and words. Based on Multi-GCN, we propose a Multiplex Graph Summarization (Multi-GraS) model for extrac- tive text summarization. Finally, we evaluate the proposed models on the CNN/DailyMail benchmark dataset to demonstrate the effec- tiveness of our method.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Journal Article:
- https://doi.org/10.3233/ICA-220680
