Turkish NLP about


I installed BotPress this week. I’m new to the project. What should I do to add Turkish NLP?

There are NLP examples, how can I position them in the project?

Hi @Ismail_Tunga,

I can’t help you integrate this repo you’ve linked.

You can however make botpress run in pretty much any language, but its a hard process.

Theses are some approximative steps to follow (for turkish):

  1. Deploy your own lang-server in local. This doc should help you.
  2. Download the sentencepiece model for turkish here. Click on “model” right next to the 200000 vocab size.
  3. Rename the dowloaded file to bp.tr.bpe.model and move it to your embeddings location. If you use OSX or Linux, this location should be ~/botpress/embeddings
  4. Dowload the fasttext model for turkish here.
  5. Rename the dowloaded file to bp.tr.300.bin and move it to the same embeddings location.
  6. Start your lang server. The language should then be available, but with no flag icon and no name.

I managed to make it work with greek models, but I don’t know why it’s not working with turkish. I believe there might be a problem with the fasttext models…

You might have to train your own fasttext models on a corpus you’ve crawled yourself…

As I said, its doable, but its a hard process…

Also, if you manage to make it work and you want to have an icon and a language name for your custom language, you can fork our repo to code the feature yourself and open a PR. All contributions are welcome :slight_smile:

I hope this information helps!

Please, right us back if you achieve making botpress run in turkish.