How can I create a dataset for italian language?

[Antonio-Nappi]

Hi, I'm new to NLP and I want to classify some text using as dataset a dump from wikipedia, how can I create such dataset and train the model on the italian language?

[elukey]

Hi Antonio! Sorry for the late answer. I would suggest to reach out to the Wikimedia ML Team via Phabricator/Libera-IRC:

https://phabricator.wikimedia.org/tag/machine-learning-team/
Libera: #wikimedia-ml

We are in the process of moving our serving infrastructure to Kubeflow, so a lot of things are changing, we'd be really happy to have a chat with you first. Adding @chrisalbon to this conversation as well (team's director).

[chrisalbon]

Hey Antonio! What kind of data are you looking for?

[Antonio-Nappi]

Hi @chrisalbon, I'd like to have Italian articles grouped by category. I would realize a topic classificator

[chrisalbon]

Hey awesome! Quarry is probably the best place to get that data. You can also use another system, PAWS, to setup your own Jupyter Notebook hosted by WMF.

[Antonio-Nappi]

Oh, thanks for the tip. Could you also help me writing the query? I suppose that, in order to have categories, I should use the "group by" clause in the query right? but how can I select all pages?