Transforming Complex Sentences into a Semantic Hierarchy

Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 3415–3427 Florence, Italy, July 28 - August 2, 2019. c⃝2019 Association for Computational Linguistics 3415 Transforming Complex Sentences into a Semantic Hierarchy Christina Niklaus13, Matthias Cetto1, Andr´e Freitas2, and Siegfried Handschuh13 1 University of St.Gallen {christina.niklaus, matthias.cetto, siegfried.handschuh}@unisg.ch 2 University of Manchester andre.freitas@manchester.ac.uk 3 University of Passau {christina.niklaus, siegfried.handschuh}@uni-passau.de Abstract We present an approach for recursively split- ting and rephrasing complex English sentences into a novel semantic hierarchy of simpli- fied sentences, with each of them present- ing a more regular structure that may facil- itate a wide variety of artificial intelligence tasks, such as machine translation (MT) or information extraction (IE). Using a set of hand-crafted transformation rules, input sen- tences are recursively transformed into a two- layered hierarchical representation in the form of core sentences and accompanying contexts that are linked via rhetorical relations. In this way, the semantic relationship of the de- composed constituents is preserved in the out- put, maintaining its interpretability for down- stream applications. Both a thorough manual analysis and automatic evaluation across three datasets from two different domains demon- strate that the proposed syntactic simplifica- tion approach outperforms the state of the art in structural text simplification. Moreover, an extrinsic evaluation shows that when applying our framework as a preprocessing step the per- formance of state-of-the-art Open IE systems can be improved by up to 346% in precision and 52% in recall. To enable reproducible re- search, all code is provided online. 1 Introduction Text Simplification (TS) is defined as the process of reducing the linguistic complexity of natural language (NL) text by utilizing a more readily ac- cessible vocabulary and sentence structure. Its goal is to improve the readability of a text, mak- ing information easier to comprehend for people with reduced literacy, such as non-native speak- ers (Paetzold and Specia, 2016), aphasics (Carroll et al., 1998), dyslexics (Rello et al., 2013) or deaf persons (Inui et al., 2003). However, not only hu- man readers may benefit from TS. Previous work has established that applying TS as a preprocess- ing step can improve the performance of a vari- ety of natural language processing (NLP) tasks, such as Open IE (Saha and Mausam, 2018; Cetto et al., 2018), MT (ˇStajner and Popovic, 2016, 2018), Relation Extraction (Miwa et al., 2010), Se- mantic Role Labeling (Vickrey and Koller, 2008), Text Summarization (Siddharthan et al., 2004; Bouayad-Agha et al., 2009), Question Generation (Heilman and Smith, 2010; Bernhard et al., 2012), or Parsing (Chandrasekar et al., 1996; Jonnala- gadda et al., 2009). Linguistic complexity stems from the use of ei- ther a difficult vocabulary or sentence structure. Therefore, TS is classified into two categories: lexical simplification and syntactic simplification. Through substituting a difficult word or phrase with a more comprehensible synonym, the for- mer primarily addresses a human audience. Most NLP systems, on the contrary, derive greater ben- efit from syntactic simplification, which focuses on identifying grammatical complexities in a sen- tence and …

Content truncated to 500 words