Sciweavers

LREC
2010

Belgisch Staatsblad Corpus: Retrieving French-Dutch Sentences from Official Documents

14 years 1 months ago
Belgisch Staatsblad Corpus: Retrieving French-Dutch Sentences from Official Documents
We describe the compilation of a large corpus of French-Dutch sentence pairs from official Belgian documents which are available in the online version of the publication Belgisch Staatsblad/Moniteur belge, and which have been published between 1997 and 2006. After downloading files in batch, we filtered out documents which have no translation in the other language, documents which contain several languages (by checking on discriminating words), and pairs of documents with a substantial difference in length. We segmented the documents into sentences and aligned the latter, which resulted in 5 million sentence pairs (only one-to-one links were included in the parallel corpus); there are 2.4 million unique pairs. Sample-based evaluation of the sentence alignment results indicates a near 100% accuracy, which can be explained by the text genre, the procedure filtering out weakly parallel articles and the restriction to one-to-one links. The corpus is larger than a number of well-known Fren...
Tom Vanallemeersch
Added 29 Oct 2010
Updated 29 Oct 2010
Type Conference
Year 2010
Where LREC
Authors Tom Vanallemeersch
Comments (0)