Hybrid text simplification using synchronous dependency grammars with hand-written and automatically harvested rules

Proceedings of the 14th Conference of the European Chapter of the Association for Computational Linguistics, pages 722–731, Gothenburg, Sweden, April 26-30 2014. c⃝2014 Association for Computational Linguistics Hybrid text simplification using synchronous dependency grammars with hand-written and automatically harvested rules Advaith Siddharthan Computing Science University of Aberdeen UK advaith@abdn.ac.uk M. A. Angrosh Computing Science University of Aberdeen UK angroshmandya@abdn.ac.uk Abstract We present an approach to text simplifi- cation based on synchronous dependency grammars. The higher level of abstraction afforded by dependency representations allows for a linguistically sound treatment of complex constructs requiring reorder- ing and morphological change, such as conversion of passive voice to active. We present a synchronous grammar formalism in which it is easy to write rules by hand and also acquire them automatically from dependency parses of aligned English and Simple English sentences. The grammar formalism is optimised for monolingual translation in that it reuses ordering infor- mation from the source sentence where ap- propriate. We demonstrate the superiority of our approach over a leading contempo- rary system based on quasi-synchronous tree substitution grammars, both in terms of expressivity and performance. 1 Introduction Text simplification is sometimes defined as the process of reducing the grammatical and lexi- cal complexity of a text, while still retaining the original information content and meaning. The main goal of simplification is to make informa- tion more accessible to the large numbers of peo- ple with reduced literacy. The National Lit- eracy Trust (http://www.literacytrust.org.uk) esti- mates that one in six adults in the UK have poor literacy skills. The situation is often worse in de- veloping countries. Alu´ısio et al. (2008) report that 68% of Brazilians between 15 and 64 years who have studied up to 4 years only reach the rudi- mentary level of literacy, and even among those who have studied for 8 years, only a quarter can be considered fully literate. While there is a large body of evidence that manual text simplification is an effective intervention (Anderson and Free- body, 1981; L’Allier, 1980; Beck et al., 1991; Anderson and Davison, 1988; Linderholm et al., 2000; Kamalski et al., 2008), there has till recently been little work on automatic simplification. The pace of research has picked up in recent years though, with many teams applying machine trans- lation approaches to perform “monolingual trans- lation” from English to simplified English. The goals of this paper are to (1) identify the limita- tions of recently published approaches to text sim- plification with regard to their coverage of linguis- tic constructs, (2) to describe an approach based on synchronous grammars operating on typed de- pendency representations that permits a more so- phisticated handling of many linguistic constructs, and (3) to present a hybrid system that combines a small set of hand written grammar rules for purely syntactic constructs with a much larger set of auto- matically acquired rules for lexicalised constructs in one synchronous formalism. We summarise work on text simplification in Section 2, before describing our method in Sec- tion 3 …

Content truncated to 500 words