Building a Monolingual Parallel Corpus for Text Simplification Using Sentence Similarity Based on Alignment between Word Embeddings

Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers, pages 1147–1158, Osaka, Japan, December 11-17 2016. Building a Monolingual Parallel Corpus for Text Simplification Using Sentence Similarity Based on Alignment between Word Embeddings Tomoyuki Kajiwara Graduate School of System Design Tokyo Metropolitan University Tokyo, Japan kajiwara-tomoyuki@ed.tmu.ac.jp Mamoru Komachi Graduate School of System Design Tokyo Metropolitan University Tokyo, Japan komachi@tmu.ac.jp Abstract Methods for text simplification using the framework of statistical machine translation have been extensively studied in recent years. However, building the monolingual parallel corpus necessary for training the model requires costly human annotation. Monolingual parallel corpora for text simplification have therefore been built only for a limited number of languages, such as English and Portuguese. To obviate the need for human annotation, we propose an unsupervised method that automatically builds the monolingual parallel corpus for text simplification using sentence similarity based on word embeddings. For any sentence pair comprising a complex sentence and its simple counterpart, we employ a many-to-one method of aligning each word in the com- plex sentence with the most similar word in the simple sentence and compute sentence similarity by averaging these word similarities. The experimental results demonstrate the excellent perfor- mance of the proposed method in a monolingual parallel corpus construction task for English text simplification. The results also demonstrated the superior accuracy in text simplification that use the framework of statistical machine translation trained using the corpus built by the proposed method to that using the existing corpora. 1 Introduction Text simplification is the process of rewriting a complex text into a simpler form while preserving its meaning. The purpose of text simplification is to assist the comprehension of readers, especially lan- guage learners and children. Recent studies have treated text simplification as a monolingual machine translation problem in which a simple synonymous sentence is generated using the framework of sta- tistical machine translation (Specia, 2010; Zhu et al., 2010; Coster and Kauchak, 2011a; Coster and Kauchak, 2011b; Wubben et al., 2012; ˇStajner et al., 2015a; ˇStajner et al., 2015b; Goto et al., 2015). However, unlike statistical machine translation, which uses bilingual parallel corpora, text simplification requires a monolingual parallel corpus for training. While bilingual parallel data are available in large quantities, monolingual parallel data are hard to obtain because simplification of a complex text is not a by-product of other tasks. Monolingual parallel corpora for text simplification are available in only seven languages—English (Zhu et al., 2010; Coster and Kauchak, 2011b; Hwang et al., 2015; Xu et al., 2015), Portuguese (Caseli et al., 2009), Spanish (Bott and Saggion, 2011), Danish (Klerke and Søgaard, 2012), German (Klaper et al., 2013), Italian (Brunato et al., 2015), and Japanese (Goto et al., 2015). In addi- tion, only the English corpora are open to the public. We therefore propose an unsupervised method 1 that automatically builds monolingual parallel corpora for text simplification without using any external resources for computing sentence similarity. In this study, a monolingual parallel corpus for text simplification is built …

Content truncated to 500 words