A pilot study on underspecified discourse connectives in the TED Talk parallel corpus
(2018) 20th DiscourseNet — Location: Budapest, Hungary (17.May.2018)
Files
No attached file found for this publication.
Details
- Authors
- et. al.
- Abstract
- Discourse connectives have two basic functions: they connect single text segments into a compound unit and they express a semantic type of this relation (contrast, reason, instantiation, etc.). In this paper we present the research methods and first findings of a multilingual study on the uses and translations of discourse connectives, in particular, connectives with underspecified functions, based on a sample of 3 TED Talks. This collection of talks can be considered as a parallel corpus of pre-planned monologues, including some dialogic elements as well, such as reports of conversations and rhetorical questions. Taking a multilingual approach we analyse not only the original English talks but their translations into Czech, French, Hungarian and Lithuanian. All occurrences of discourse connectives (such as and, but, so, and when) along with their translations were manually identified and annotated in each language under scrutiny by two expert annotators, following Crible & Degand’s (in press) functional classification, which we have found the most suitable for the analysis of spoken language. Inter-annotator disagreements were discussed and resolved via annotators’ meetings by applying a gold standard, i.e. reaching the consensual. Our group has a threefold objective: (1) to identify discourse connectives and delimit the boundaries of this fuzzy category across languages; (2) to assign functions to these connectives, and (3) to compare the typical occurrences of ambiguity and underspecification in different languages. Based on our initial observations, we have noted several interesting phenomena related to fuzziness/underspecification: (1) implicitation in translation (e.g. “now” not translated in other languages, or connectives with sequential functions omitted following the requirements of synchronizing the subtitles, while rhetorical and ideational functions preserved in the corresponding translations), (2) “underspecification” in translation (e.g. “although” translated into “but”), and (3) monolingual underspecification (e.g. “and” used to express contrast, whether or not it is the original or the translated text). On the whole, the analysis of the multilingual data shows that some processes of the underspecification are general and occur in many languages in parallel. This concerns especially systematic shifts between discourse domains (ideational, rhetorical, sequential, interpersonal, cf. Crible and Degand, in press) which are most typical for spoken language.
- Affiliations
Citations
Abuzcki, A., Burkšaitienė, N., Crible, L., & et al. (2018). A pilot study on underspecified discourse connectives in the TED Talk parallel corpus. 20th DiscourseNet, Budapest, Hungary. https://hdl.handle.net/2078.5/176570