On the Similarities Between Native, Non-native and Translated Texts

Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics, pages 1870–1881, Berlin, Germany, August 7-12, 2016. c⃝2016 Association for Computational Linguistics On the Similarities Between Native, Non-native and Translated Texts Ella Rabinovich△⋆ Sergiu Nisioi⋄ Noam Ordan† Shuly Wintner⋆ △IBM Haifa Research Labs ⋆Department of Computer Science, University of Haifa ⋄Solomon Marcus Center for Computational Linguistics, University of Bucharest †The Arab College for Education, Haifa {ellarabi,sergiu.nisioi,noam.ordan}@gmail.com, shuly@cs.haifa.ac.il Abstract We present a computational analysis of three language varieties: native, advanced non-native, and translation. Our goal is to investigate the similarities and differ- ences between non-native language pro- ductions and translations, contrasting both with native language. Using a collec- tion of computational methods we estab- lish three main results: (1) the three types of texts are easily distinguishable; (2) non- native language and translations are closer to each other than each of them is to native language; and (3) some of these character- istics depend on the source or native lan- guage, while others do not, reflecting, per- haps, unified principles that similarly af- fect translations and non-native language. 1 Introduction This paper addresses two linguistic phenomena: translation and non-native language. Our main goal is to investigate the similarities and differ- ences between these two phenomena, and contrast them with native language. In particular, we are interested in the reasons for the differences be- tween translations and originals, on one hand, and native and non-native language, on the other. Do they reflect “universal” principles, or are they de- pendent on the source/native language? Much research in translation studies indicates that translated texts have unique characteristics. Translated texts (in any language) constitute a sub- language of the target language, sometimes re- ferred to as translationese (Gellerstam, 1986). The unique characteristics of translationese have been traditionally classified into two categories: proper- ties that stem from interference of the source lan- guage (Toury, 1979), and universal traits resulting from the translation process itself, independently of the specific source and target languages (Baker, 1993; Toury, 1995). The latter so-called transla- tion universals have triggered a continuous debate among translation studies researchers (Mauranen and Kujam¨aki, 2004; House, 2008; Becher, 2010). Similarly, over half a century of research on second language acquisition (SLA) established the presence of cross-linguistic influences (CLI) in non-native utterances (Jarvis and Pavlenko, 2008). CLI is a cover term proposed by Kellerman and Sharwood-Smith (1986) to denote various phe- nomena that stem from language contact situa- tions such as transfer, interference, avoidance, bor- rowing, etc.1 In addition, universal traits result- ing from the learning process itself have been no- ticed regardless of the native language, L1.2 For example, similar developmental sequences have been observed for negation, question formation, and other sentence structures in English (Dulay and Burt, 1974; Odlin, 1989) for both Chinese and Spanish natives. Phenomena such as overgener- alization, strategies of learning (Selinker, 1972), psychological factors (Ellis, 1985), and cultural distance (Giles and Byrne, 1982) are also influen- tial in the acquisition process. There are clear similarities between translations and non-native …

Content truncated to 500 words