Cohesion in subtitles: A corpus-based study

Abstract

It has been shown in earlier research on comparable corpora that translated language differs from non-translated language in its use of the linguistic resources of the target language. For example, language-specific or “unique” items have been shown to manifest significantly lower frequencies in translated Finnish. There are reasons to expect, however, that subtitle language might differ from other varieties of translated language in its exploitation of such cohesive devices that contribute to brevity and conciseness. It is now possible to test this hypothesis on a subtitle corpus of 100 million words which we have recently compiled from the subtitle files of the Finnish Broadcasting Company (FBC/YLE). The clitic particles and other short cohesive elements are an excellent case in point for testing the hypothesis. This paper will report on our preliminary findings of the cohesive devices as they are used in subtitled Finnish of the FBC corpus. Our results suggest that the “unique” cohesive particles manifest even greater frequencies in subtitles than in varieties of non-translated language. Thus it seems that subtitles do not conform to some of the patterns found in previous research on translated language.

The VIW project

“…Still in the field of subtitling, other interesting resources are the ESIST corpus, which consists of 48 subtitled versions of three short segments developed within the Comparative Subtitling Project (www.esist.org/comparative-subtitling-project/). And the corpus compiled by Tirkkonen-Condit and Mäkisalo (2007) from the Finnish Broadcasting Company subtitle files, totalling more than 100 million words, which has been analysed by the authors to identify, for instance, cohesive devices. A final example is the multilingual corpus created by Mouka et al (2012), including five films in English with English,…”

Concept Selection and Translation Strategy: Subtitling for the Deaf based on Corpus Analysis

“…Interlingual subtitling has been studied with a corpus, though in most cases the corpus was not annotated (Mattsson, 2009;Tiedemann, 2007;Tirkkonen-Condit & Mäkisalo, 2007). The Veiga English-Galician corpus (Sotelo Dios & Gómez Guinovart, 2012) is composed of 40 audiovisual texts, which are partially annotated.…”