UEdin participation in the 1st Translation Memory Cleaning Shared Task

Christian Buck, Philipp Koehn

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract / Description of output

We present our submission for the 1st Translation Memory Cleaning Shared Task. We treat the task as a 3-class classification problem and extract features that indicate (i) source sentence complexity, (ii) misalignments between source and target, and (iii) target sentence complexity. Our results show that focusing on the target side and finding ways to estimate the alignment quality between source and target yields expressive features which, together with a reliable classifier, produces competitive results. Our submission is ranked on 2nd place among 6 for the EN-DE language pair
Original languageEnglish
Title of host publicationProceedings of 2nd Workshop on Natural Language Processing for Translation Memories (NLP4TM 2016)
Place of PublicationPortorož, Slovenia
Number of pages4
Publication statusPublished - May 2016
Event2nd Workshop on Natural Language Processing for Translation Memories - Portorož, Slovenia
Duration: 28 May 201628 May 2016
http://rgcl.wlv.ac.uk/nlp4tm2016/

Conference

Conference2nd Workshop on Natural Language Processing for Translation Memories
Abbreviated titleNLP4TM 2016
Country/TerritorySlovenia
CityPortorož
Period28/05/1628/05/16
Internet address

Fingerprint

Dive into the research topics of 'UEdin participation in the 1st Translation Memory Cleaning Shared Task'. Together they form a unique fingerprint.

Cite this