Improving Translation via Targeted Paraphrasing

Proceedings of the 2010 Conference on Empirical Methods in Natural Language Processing, pages 127–137, MIT, Massachusetts, USA, 9-11 October 2010. c⃝2010 Association for Computational Linguistics Improving Translation via Targeted Paraphrasing Philip Resnik Linguistics and UMIACS University of Maryland resnik@umd.edu Olivia Buzek Linguistics and Computer Science University of Maryland olivia.buzek@gmail.com Chang Hu Computer Science University of Maryland changhu@cs.umd.edu Yakov Kronrod Linguistics and UMIACS University of Maryland yakov@umd.edu Alex Quinn Computer Science University of Maryland aq@cs.umd.edu Benjamin B. Bederson Computer Science and UMIACS University of Maryland bederson@cs.umd.edu Abstract Targeted paraphrasing is a new approach to the problem of obtaining cost-effective, reasonable quality translation that makes use of simple and inexpensive human computations by monolin- gual speakers in combination with machine translation. The key insight behind the process is that it is possible to spot likely translation errors with only monolingual knowledge of the target language, and it is possible to generate al- ternative ways to say the same thing (i.e. para- phrases) with only monolingual knowledge of the source language. Evaluations demon- strate that this approach can yield substantial improvements in translation quality. 1 Introduction For most of the world’s languages, the availability of translation is limited to two possibilities: high qual- ity at high cost, via professional translators, and low quality at low cost, via machine translation (MT). The spectrum between these two extremes is very poorly populated, and at any point on the spectrum the ready availability of translation is limited to only a small fraction of the world’s languages. There is, of course, a long history of technological assistance to transla- tors, improving cost effectiveness using translation memory (Laurian, 1984; Bowker and Barlow, 2004) or other interactive tools to assist translators (Esteban et al., 2004; Khadivi et al., 2006). And there is a recent and rapidly growing interest in crowdsourc- ing with non-professional translators, which can be remarkably effective (Munro, 2010). However, all these alternatives face a central availability bottle- neck: they require the participation of humans with bilingual expertise. In this paper, we report on a new exploration of the middle ground, taking advantage of a virtually unutilized resource: speakers of the source and tar- get language who are effectively monolingual, i.e. who each only know one of the two languages rel- evant for the translation task. The solution we are proposing has the potential to provide a more cost effective approach to translation in scenarios where machine translation would be considered acceptable to use, if only it were generally of high enough qual- ity. This would clearly exclude tasks like transla- tion of medical reports, business contracts, or literary works, where the validation of a qualified bilingual translator is absolutely necessary. However, it does include a great many real-world scenarios, such as following news reports in another country, reading in- ternational comments about a product, or generating a decent first draft translation of a Wikipedia page for Wikipedia editors to improve. The use of monolingual participants in a human- machine translation process is not entirely new. Callison-Burch …

Content truncated to 500 words