Discourse connectives in translation

Discourse connectives, for example because, although, or moreover, explicitly mark the coherence relations that hold between segments in a text. When a text is translated into another language, its overall meaning, including how different parts of the text relate to each other, is, ideally, preserved. Since connectives are often reformulated, added, or removed in translation (Halverson 2004, Zufferey & Cartoni 2014), supposedly without detrimental effects on the text as a whole, it is worth examining when variation in connective translation occurs and what the limitations of this variation are. The way in which connectives are translated cannot only provide valuable insight into the differences in linguistic resources between two languages in a translation pair, but also into the nature of discourse coherence in general and the meaning of connectives in particular. In this study, we annotated a set of English coherence relations (n≈2000) taken from the Europarl corpus (Cartoni, Zufferey, & Meyer 2013, Koehn 2005), marked by also, although, because, but, if, in addition, so, and unless, using an annotation scheme based on Sanders, Spooren, and Noordman (1992). In addition to annotating the type of relation the connectives signaled, we determined for each relation whether it was embedded under a syntactic construction, as in (1a), and whether it was involved in another explicit coherence relation, for instance if it was embedded in another relation, as the conditional relation in (1b). (1) a. I am sad he won because the competition was disqualified. Syntactic dependency b. I want you to get dressed, because if we do not leave the house soon we are going to be late. Relational dependency We subsequently located the equivalent of each connective in the Dutch, French, German, and Spanish translations, and analyzed them using five categories: explicit, paraphrase, underspecified connective, syntax, and implicit. These categories are illustrated by the simplified examples in (2a-e). (2) a. Because John won the race, he is very happy. Explicit b. John’s victory made him very happy. Paraphrase c. John won the race and was very happy. Underspecified connective d. John, who won the race, was very happy. Syntax e. John won the race. Ø He was very happy. Implicit We are currently in the process of finishing up our corpus annotation. We aim to model the influence of the type of relation, syntactic dependency, and relational dependency on the way in which the relations are translated. We will both look at the categories outlined in (2), and, if the translation features a connective (which is the case for the majority of relations), the specific connective used in the translation. Preliminary conclusions are that the type of relation influences whether the relation is also explicitly signaled in the translation, that syntactic dependency influences the choice of one connectives over others (e.g., syntactically embedded becauses are predominantly translated by omdat in Dutch), and that there seems to be a tendency across language pairs to break up connectives that signal highly complex relations (unless). References Cartoni, B., Zufferey, S., & Meyer, T. (2013). Using the Europarl corpus for linguistics research. Belgian Journal of Linguistics 27, 23-42. Halverson, S. (2004). Connectives as a translation problem. In: H. Kittel et al. (Eds.), An International Encyclopedia of Translation Studies, 562¬–572. Berlin/New York: Walter de Gruyter. Sanders T.J.M., Spooren W.P.M.S. & Noordman L.G.M. (1992). Towards a Taxonomy of Coherence Relations. Discourse Processes 15, 1–36. Zufferey, S. & Cartoni, B. (2014). A multifactorial analysis of explicitation in translations. Target, 26, 361–384.