Abstractive Text Summarization Based on Deep Learning and Semantic Content Generalization

Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 5082–5092 Florence, Italy, July 28 - August 2, 2019. c⃝2019 Association for Computational Linguistics 5082 Abstractive text summarization based on deep learning and semantic content generalization Panagiotis Kouris School of Electrical & Computer Engineering National Technical University of Athens, Greece pkouris@islab.ntua.gr Georgios Alexandridis School of Electrical & Computer Engineering National Technical University of Athens, Greece gealexandri@islab.ntua.gr Andreas Stafylopatis School of Electrical & Computer Engineering National Technical University of Athens, Greece andreas@cs.ntua.gr Abstract This work proposes a novel framework for en- hancing abstractive text summarization based on the combination of deep learning tech- niques along with semantic data transfor- mations. Initially, a theoretical model for semantic-based text generalization is intro- duced and used in conjunction with a deep encoder-decoder architecture in order to pro- duce a summary in generalized form. Sub- sequently, a methodology is proposed which transforms the aforementioned generalized summary into human-readable form, retaining at the same time important informational as- pects of the original text and addressing the problem of out-of-vocabulary or rare words. The overall approach is evaluated on two pop- ular datasets with encouraging results. 1 Introduction Text Summarization (TS) aims at composing a concise version of an original text, retaining its salient information. Since manual TS is a de- manding, time expensive and generally laborious task, automatic TS is gaining increasing popular- ity and therefore constitutes a strong motivation for further research. Current efforts in automatic TS mainly focus on summarizing single documents (e.g. news, ar- ticles, scientific papers, weather forecasts, etc.) and multi-documents (e.g. news from different sources, user reviews, e-mails etc.), reducing the size of the initial text while at the same time pre- serving key informational elements and the mean- ing of content. Two main approaches to automatic TS have been reported in the relevant literature; extractive and abstractive (Gambhir and Gupta, 2017; Allah- yari et al., 2017). In the former case, those sen- tences of original text that convey its content are firstly identified and then extracted in order to con- struct the summary. In the latter case, new sen- tences are generated which concatenate the over- all meaning of the initial text, rephrasing its con- tent. Abstractive TS is a more challenging task; it resembles human-written summaries, as it may contain rephrased sentences or phrases with new words (i.e. sentences, phrases and words that do not appear in the original text), thereby improv- ing the generated summary in terms of cohesion, readability or redundancy. The main contribution of this work is a novel abstractive TS technique that combines deep learning models of encoder-decoder architecture and semantic-based data transformations. Since the majority of literature in abstractive TS focuses in either of the aforementioned parts, the proposed approach tries to bridge this gap by introducing a framework that combines the potential of machine learning with the importance of semantics. The said framework is comprised of three components; (i) a theoretical model for text generalization (Sec- tion 3) (ii) a …

Content truncated to 500 words