If you can read this, your browser provides insufficient support for style sheets. The visual presentation of this document will suffer.
 

Pado and Erk 2010

S. Pado and K. Erk. Translation Shifts and Frame-Semantic Mismatches: A Corpus Analysis. International Journal of Corpus Linguistics. Accepted for publication. Contact me for preprint.

Translation shifts form an important area of research in translation studies, and knowledge about shifts can be beneficial for both practical translation and multilingual language processing. However, manual quantitative assessment of shifts is quite expensive. At the same time, the semantic analysis of text in terms of word sense and semantic roles is an active research topic in computational linguistics. Since semantic roles reflect predicate-argument structure, it is an attractive hypothesis that cross-lingual mismatches of sense and semantic role annotation in a parallel corpus can be interpreted as (semantic) translation shifts.

This article reports on a study testing this hypothesis on a 1000-sentence parallel corpus (English--German) with FrameNet frames and semantic roles. The results are mixed: A significant portion, but overall a minority, of cross-lingual semantic mismatches indicate semantic translation shifts. We identify two major sources of spurious mismatches. The first one is annotator error, which may be eliminated in future studies. The second one is more fundamental, namely the difficulty to formulate criteria that consistently distinguish between semantic categories. 


@Article{pado07:_depen,
  author = 	 {Sebastian Pad\'o and Katrin Erk},
  title = 	 {Translation Shifts and Frame-Semantic Mismatches: A Corpus Analysis},
  journal = 	 {International Journal of Corpus Linguistics},
  year = 	 2010,
  note = {To appear}
}