Simultaneous Interpreting as a Noisy Channel: How Much Information Gets Through

Proceedings of Recent Advances in Natural Language Processing, pages 608–618 Varna, Sep 4–6, 2023 https://doi.org/10.26615/978-954-452-092-2_066 608 Simultaneous Interpreting as a Noisy Channel: How Much Information Gets Through Maria Kunilovskaya Saarland University, Saarbr¨ucken maria.kunilovskaya@uni-saarland.de Heike Przybyl Saarland University, Saarbr¨ucken heike.przybyl@uni-saarland.de Elke Teich Saarland University, Saarbr¨ucken e.teich@mx.uni-saarland.de Ekaterina Lapshinova-Koltunski University of Hildesheim lapshinovakoltun@uni-hildesheim.de Abstract We explore the relationship between infor- mation density/surprisal of source and tar- get texts in translation and interpreting in the language pair English-German, looking at the specific properties of translation (“transla- tionese”). Our data comes from two bidirec- tional English-German subcorpora represent- ing written and spoken mediation modes col- lected from European Parliament proceedings. Within each language, we (a) compare origi- nal speeches to their translated or interpreted counterparts, and (b) explore the association between segment-aligned sources and targets in each translation direction. As additional variables, we consider source delivery mode (read-out, impromptu) and source speech rate in interpreting. We use language modelling to measure the information rendered by words in a segment and to characterise the cross-lingual transfer of information under various condi- tions. Our approach is based on statistical analyses of surprisal values, extracted from n- gram models of our dataset. The analysis re- veals that while there is a considerable positive correlation between the average surprisal of source and target segments in both modes, in- formation output in interpreting is lower than in translation, given the same amount of in- put. Significantly lower information density in spoken mediated production compared to non- mediated speech in the same language can in- dicate a possible simplification effect in inter- preting. 1 Introduction In this study, we describe and explain linguistic choice in translation and interpreting from the point of view of rational communication, according to which language users strive to encode their mes- sages effectively and efficiently, i.e. they attempt to ensure that their messages are transmitted success- fully while at the same time, their cognitive effort stays at a reasonable level (see e.g. Crocker et al., 2015). Our approach stipulates that the behaviour of translators, while guided by effectiveness and efficiency, is severely constrained by the specific conditions of mediated communication, especially in interpreting (see studies on the cognitive effort in interpreting, e.g. Christoffels et al., 2006; Chmiel, 2021). Simultaneous interpreters have to balance allocating cognitive resources to overlapping com- prehension and production processes in a way that allows them to complete the task and communica- tion is not put at risk. From empirical translatology we know that the coping mechanisms involved in transla- tion/interpreting have an impact on the linguistic properties of the output, widely known as transla- tionese (e.g. Baker, 1996; Teich, 2003; Shlesinger and Ordan, 2012, cf. Section 2). While there is a rich literature on trends in translational behaviour (e.g. simplification, explicitation, normalisation), a unifying explanation for the diverse linguistic phe- nomena is still lacking. This study is an attempt to fill this gap by adopting an information-theoretic approach. Our analysis is based on measuring information density (ID) aka surprisal of transla- …

Content truncated to 500 words