Hybrid Simplification using Deep Semantics and Machine Translation

Proceedings of the 52nd Annual Meeting of the Association for Computational Linguistics, pages 435–445, Baltimore, Maryland, USA, June 23-25 2014. c⃝2014 Association for Computational Linguistics Hybrid Simplification using Deep Semantics and Machine Translation Shashi Narayan Universit´e de Lorraine, LORIA Villers-l`es-Nancy, F-54600, France shashi.narayan@loria.fr Claire Gardent CNRS, LORIA, UMR 7503 Vandoeuvre-l`es-Nancy, F-54500, France claire.gardent@loria.fr Abstract We present a hybrid approach to sentence simplification which combines deep se- mantics and monolingual machine transla- tion to derive simple sentences from com- plex ones. The approach differs from pre- vious work in two main ways. First, it is semantic based in that it takes as in- put a deep semantic representation rather than e.g., a sentence or a parse tree. Sec- ond, it combines a simplification model for splitting and deletion with a monolin- gual translation model for phrase substi- tution and reordering. When compared against current state of the art methods, our model yields significantly simpler out- put that is both grammatical and meaning preserving. 1 Introduction Sentence simplification maps a sentence to a sim- pler, more readable one approximating its con- tent. Typically, a simplified sentence differs from a complex one in that it involves simpler, more usual and often shorter, words (e.g., use instead of exploit); simpler syntactic constructions (e.g., no relative clauses or apposition); and fewer mod- ifiers (e.g., He slept vs. He also slept). In prac- tice, simplification is thus often modeled using four main operations: splitting a complex sen- tence into several simpler sentences; dropping and reordering phrases or constituents; substituting words/phrases with simpler ones. As has been argued in previous work, sentence simplification has many potential applications. It is useful as a preprocessing step for a variety of NLP systems such as parsers and machine trans- lation systems (Chandrasekar et al., 1996), sum- marisation (Knight and Marcu, 2000), sentence fusion (Filippova and Strube, 2008) and semantic role labelling (Vickrey and Koller, 2008). It also has wide ranging potential societal application as a reading aid for people with aphasis (Carroll et al., 1999), for low literacy readers (Watanabe et al., 2009) and for non native speakers (Siddharthan, 2002). There has been much work recently on de- veloping computational frameworks for sentence simplification. Synchronous grammars have been used in combination with linear integer program- ming to generate and rank all possible rewrites of an input sentence (Dras, 1999; Woodsend and La- pata, 2011). Machine Translation systems have been adapted to translate complex sentences into simple ones (Zhu et al., 2010; Wubben et al., 2012; Coster and Kauchak, 2011). And handcrafted rules have been proposed to model the syntactic transformations involved in simplifications (Sid- dharthan et al., 2004; Siddharthan, 2011; Chan- drasekar et al., 1996). In this paper, we present a hybrid approach to sentence simplification which departs from this previous work in two main ways. First, it combines a model encoding probabil- ities for splitting and deletion with a monolin- gual machine translation module which handles reordering and substitution. In this way, we ex- ploit the ability of statistical …

Content truncated to 500 words