Multext corpora

Two different corpora have been used for developping linguistic tools within MULTEXT project :

• 1. a set of pieces from the Official Journal of European Community (JOC) in five languages (English, German, Italian Spanish and French);
  The JOC corpus is delivered in Corpus Encoding Standard conformant format at each level of treatment :
  
  
  • first level : paragraph annotation level, conformant to CESDOC specifications (40 files in five languages);

  • second level : morpho-syntactic annotation level (PoS Tagging), conformant to CESANA specifications (10 files in five languages);

  • third level : parallel text alignment level, conformant to CESALIGN specifications (10 files in four pairs of languages).

• 2. a comparable corpus of Swedish financial newstext (Dagens Industri 1993 : 14 180 articles from the Swedish daily financial newspaper Dagens Industri), provided in four versions :

  • initial raw data (274 files);

  • tokenized data (274 files);

  • tokenized data with typographically generated structural markup (274 files);

  • result of a sequence of fully automatic processing steps, consisting of tokenization, structural markup, lexical lookup and probabilistic disambiguation (12 files).

  The Swedish Multext Comparable Corpus is completed by the Department of Linguistics of the University of Umea.

You are invited to send comments and feedback to multext@lpl.univ-aix.fr.