As a framework for extractive summarization, sentence regression\u00a0has achieved state-of-the-art performance in several widely-used\u00a0practical systems. The most challenging task within the sentence\u00a0regression framework is to identify discriminative features to encode\u00a0a sentence into a feature vector. So far, sentence regression\u00a0approaches have neglected to use features that capture contextual\u00a0relations among sentences.\u00a0We propose a neural network model, Contextual Relation-based\u00a0Summarization (CRSum), to take advantage of contextual relations\u00a0among sentences so as to improve the performance of sentence\u00a0regression. Specifically, we first use sentence relations with a wordlevel\u00a0attentive pooling convolutional neural network to construct\u00a0sentence representations. Then, we use contextual relations with a\u00a0sentence-level attentive pooling recurrent neural network to construct\u00a0context representations. Finally, CRSum automatically learns\u00a0useful contextual features by jointly learning representations of\u00a0sentences and similarity scores between a sentence and sentences\u00a0in its context. Using a two-level attention mechanism, CRSum is\u00a0able to pay attention to important content, i.e., words and sentences,\u00a0in the surrounding context of a given sentence.\u00a0We carry out extensive experiments on six benchmark<\/p>\n
We propose a neural network model, Contextual Relation-based\u00a0Summarization (CRSum), to take advantage of contextual relations\u00a0among sentences so as to improve the performance of sentence\u00a0regression. Specifically, we first use sentence relations with a wordlevel\u00a0attentive pooling convolutional neural network to construct\u00a0sentence representations. Then, we use contextual relations with a\u00a0sentence-level attentive pooling recurrent neural network to construct\u00a0context representations. Finally, CRSum automatically learns\u00a0useful contextual features by jointly learning representations of\u00a0sentences and similarity scores between a sentence and sentences\u00a0in its context. Using a two-level attention mechanism, CRSum is\u00a0able to pay attention to important content, i.e., words and sentences,\u00a0in the surrounding context of a given sentence.<\/p>\n
We carry out extensive experiments on six benchmark datasets.\u00a0CRSum alone can achieve comparable performance with state-of-the-art\u00a0approaches; when combined with a few basic surface features,\u00a0it signi\u0080cantly outperforms the state-of-the-art in terms of\u00a0multiple ROUGE metrics.<\/p>\n","protected":false},"excerpt":{"rendered":"
As a framework for extractive summarization, sentence regression\u00a0has achieved state-of-the-art performance in several widely-used\u00a0practical systems. The most challenging task within the sentence\u00a0regression framework is to identify discriminative features to encode\u00a0a sentence into a feature vector. So far, sentence regression\u00a0approaches have neglected to use features that capture contextual\u00a0relations among sentences.\u00a0We propose a neural network model, Contextual […]<\/p>\n","protected":false},"featured_media":0,"template":"","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"footnotes":""},"msr-content-type":[3],"msr-research-highlight":[],"research-area":[13555],"msr-publication-type":[193716],"msr-product-type":[],"msr-focus-area":[],"msr-platform":[],"msr-download-source":[],"msr-locale":[268875],"msr-field-of-study":[],"msr-conference":[],"msr-journal":[],"msr-impact-theme":[],"msr-pillar":[],"class_list":["post-400478","msr-research-item","type-msr-research-item","status-publish","hentry","msr-research-area-search-information-retrieval","msr-locale-en_us"],"msr_publishername":"ACM","msr_edition":"SIGIR \u201917, Shinjuku, Tokyo, Japan, August 07-11, 2017","msr_affiliation":"","msr_published_date":"2017-08-07","msr_host":"","msr_duration":"","msr_version":"","msr_speaker":"","msr_other_contributors":"","msr_booktitle":"","msr_pages_string":"","msr_chapter":"","msr_isbn":"","msr_journal":"","msr_volume":"","msr_number":"","msr_editors":"","msr_series":"","msr_issue":"","msr_organization":"","msr_how_published":"","msr_notes":"","msr_highlight_text":"","msr_release_tracker_id":"","msr_original_fields_of_study":"","msr_download_urls":"","msr_external_url":"","msr_secondary_video_url":"","msr_longbiography":"","msr_microsoftintellectualproperty":1,"msr_main_download":"400481","msr_publicationurl":"","msr_doi":"","msr_publication_uploader":[{"type":"file","title":"ren-leveraging-2017","viewUrl":"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2017\/07\/ren-leveraging-2017.pdf","id":400481,"label_id":0}],"msr_related_uploader":"","msr_attachments":[],"msr-author-ordering":[{"type":"text","value":"Pengjie Ren","user_id":0,"rest_url":false},{"type":"text","value":"Zhumin Chen","user_id":0,"rest_url":false},{"type":"text","value":"Zhaochun Ren","user_id":0,"rest_url":false},{"type":"user_nicename","value":"fuwei","user_id":31830,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=fuwei"},{"type":"text","value":"Jun Ma","user_id":0,"rest_url":false},{"type":"text","value":"Maarten de Rijke","user_id":0,"rest_url":false}],"msr_impact_theme":[],"msr_research_lab":[],"msr_event":[],"msr_group":[],"msr_project":[],"publication":[],"video":[],"download":[],"msr_publication_type":"inproceedings","_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/400478"}],"collection":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item"}],"about":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/types\/msr-research-item"}],"version-history":[{"count":1,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/400478\/revisions"}],"predecessor-version":[{"id":400484,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/400478\/revisions\/400484"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=400478"}],"wp:term":[{"taxonomy":"msr-content-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-content-type?post=400478"},{"taxonomy":"msr-research-highlight","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-highlight?post=400478"},{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=400478"},{"taxonomy":"msr-publication-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-publication-type?post=400478"},{"taxonomy":"msr-product-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-product-type?post=400478"},{"taxonomy":"msr-focus-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-focus-area?post=400478"},{"taxonomy":"msr-platform","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-platform?post=400478"},{"taxonomy":"msr-download-source","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-download-source?post=400478"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=400478"},{"taxonomy":"msr-field-of-study","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-field-of-study?post=400478"},{"taxonomy":"msr-conference","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-conference?post=400478"},{"taxonomy":"msr-journal","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-journal?post=400478"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=400478"},{"taxonomy":"msr-pillar","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-pillar?post=400478"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}