7

The brown corpus is a collection of text where each element is already gramatically tagged. It contains about one million words and is often used to train statistical PoS taggers.

For the french language, I'm aware of two equivalents :

However, I can't use the first one for licensing reasons and the latter is too small for my purpose (3000 sentences).

Is there an already existing dataset satisfying the following conditions?

  • French language
  • texts from various sources (newspaper, novels,...)
  • precisely tagged (at least sentences and words)
  • a size (at least) similar to the brown corpus
1

Via Twitter, I asked a friend with expertise in computational linguistics and French. She stated that the French Treebank is the largest tagged corpus. Since both it and the Brown corpus are described as about 1M words, I don't think you'll find another French one which meets your final condition.

It also seems, if I understand correctly, that the French Treebank is all sourced from Le Monde, so even it fails the second condition ("various sources").

Ultimately, her question was why the French Treebank can't be licensed, as the answer might help to make a better recommendation.

0

The French originals of the Europarl Corpus (https://en.wikipedia.org/wiki/Europarl_corpus and Links therein) may satisfy your requirements. It is however composed of only one genre (parliamentary speeches), so the variation requirement is not met.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.