The ALLEGRA corpus for German, Romansh and Italian

Description

The ALLEGRA corpus is a new language resource for Rumantsch Grischun. It takes the form of a sentence-aligned trilingual corpus consisting of press releases in the three official languages of the canton of Grisons (its name is the acronym for “ALigned press reLEases of the GRisons Administration”; allegra also means 'hello' in Romansh). The web site of the Grisons administration gives access to all press releases since 1997. Most of these releases have been written in German and translated to Rumantsch Grischun and Italian.

The ALLEGRA corpus was prepared as follows:

• All documents were downloaded, cleaned and converted to plain text format.
• Only documents available in the three languages were kept.
• Until 2009, the different language versions of a document were not linked. This linkage was added manually on the basis of the release date and the title of the document.
• All documents were sentence-aligned with a standard alignment tool (Gale and Church, 1993).

Reference

The ALLEGRA corpus is described in more detail in our LREC 2012 paper:

Yves Scherrer & Bruno Cartoni (2012). The Trilingual ALLEGRA Corpus: Presentation and Possible Use for Lexicon Induction. Proceedings of the Eighth International Conference on Language Resources and Evaluation (LREC 2012), Istanbul.
PDF

Downloads

• Raw corpus, aligned at document level (all documents with the same file name are translations of each other).
  ZIP, 10MB
• Sentence-aligned corpus. No further pre-processing (tokenization, filtering, lowercasing, stemming...) was performed.
  ZIP, 14MB

Yves Scherrer & Bruno Cartoni, 7.5.2012