0

I would like to know if there are Stanford CoreNLP models for named entity recognition and relation extraction for French text. If yes, where can I download them? Thanks.

Helena Galhardas
  • 1

1 Answers1

2

At the moment we don't distribute a French NER model, because we don't have French NER data.

But I looked around and found this data set:

https://github.com/EuropeanaNewspapers/ner-corpora

So potentially I can train a model off of it.

If you don't want to wait for me to train a model, there are instructions on this page for training a model on your own!

https://nlp.stanford.edu/software/crf-faq.html#a

StanfordNLPHelp
  • 8,699
  • 1
  • 11
  • 9
  • https://stackoverflow.com/questions/37852084/how-to-train-a-french-ner-based-on-stanford-nlp-conditional-random-fields-model – StanfordNLPHelp Nov 03 '17 at 02:48
  • I looked at the Europeana link you sent (thanks!). As far as I understand, it has annotations for entities but not for relations. So, if I want to extract both, entities and relations, from French text, this resource wouldn't be enough to train models able to do that. Do you confirm? – Helena Galhardas Nov 07 '17 at 22:23
  • Yes, I am unaware of any French relation data sets out there. You might get decent results by translating French sentences to English (Google and Facebook have high quality translation API's, though not perfect) and using English relation extraction on the translated sentence. – StanfordNLPHelp Nov 08 '17 at 01:02
  • I should note that when we built our Spanish relation extraction system, we didn't use a statistical model and basically translated the rule-based system from English to Spanish. The rule-based system alone is not that much worse than statistical systems. – StanfordNLPHelp Nov 08 '17 at 01:04