A Corpus of Native, Non-native and Translated Texts

A Corpus of Native, Non-native and Translated Texts Sergiu Nisioi⋄, Ella Rabinovich∗, Liviu P. Dinu⋄, Shuly Wintner∗ ⋄Center for Computational Linguistics, University of Bucharest ∗Department of Computer Science, University of Haifa sergiu.nisioi@gmail.com, ellarabi@csweb.haifa.ac.il, ldinu@fmi.unibuc.ro, shuly@cs.haifa.ac.il Abstract We describe a monolingual English corpus of original and (human) translated texts, with an accurate annotation of speaker properties, including the original language of the utterances and the speaker’s country of origin. We thus obtain three sub-corpora of texts reflecting native English, non-native English, and English translated from a variety of European languages. This dataset will facilitate the investi- gation of similarities and differences between these kinds of sub-languages. Moreover, it will facilitate a unified comparative study of translations and language produced by (highly fluent) non-native speakers, two closely-related phenomena that have only been studied in isolation so far. Keywords: Corpus linguistics, Translation, Bilingualism, Second language acquisition 1. Introduction Interlanguage is the entire linguistic system that emerges when second language learners express meaning in the tar- get language (Selinker, 1972). One of the main character- istics of this system are the so-called cross-linguistic influ- ences - a cover term proposed by Kellerman and Sharwood- Smith (1986) to denote various phenomena that stem from language contact situations such as transfer, interference, avoidance, borrowing, etc. Language transfer, in particular, is the influence resulting from similarities and differences between the target language (L2) and any other language that has been previously acquired (L1) (Odlin, 1989). The centerpiece of interlanguage resides in the concept of fossilization, i.e., “the permanent cessation of target lan- guage learning before the learner has attained the L2 norms at all levels of linguistic structure” (Selinker, 1972; Han, 2013). In other words, the language learning process rarely (Lardiere, 2006) leads to a full, or native-like acquisition of the target language. Under this assumption, non-native speakers are distinguishable from native speakers, no mat- ter how proficient they are in the target language. However, researchers (Birdsong, 1992; White and Genesee, 1996; Long, 2003) indicate that fossilization does not occur on all levels of linguistic structure, rather stabilization might occur on certain levels, while others continue to develop. Toury (1979) claimed that interlanguage is not only present in the context of non-native speakers, but also, to a cer- tain degree, in translated texts, which presumably reflect both the artifacts of the translation process and traces of the original language. Research in translation studies (Fraw- ley, 1984; Baker, 1996) indicates that translated texts have unique characteristics. Gellerstam (1986) suggested that the differences between original and translated texts do not indicate poor translation but rather a statistical phe- nomenon, caused by a systematic transfer of the source lan- guage into the target one. Specific phenomena that characterize the contact between two languages in non-native utterances or translated texts have so far been studied in isolation, both in the linguistic literature and in terms of computational investigations. In this paper we describe a corpus constructed from original English utterances (where we differentiate between native and non-native speakers based on their …

Content truncated to 500 words