Will the corpus be openly published?

[avacaondata]

Given how good this model works, I'd be interested in having access to the original corpus with which it was trained. Will that be possible? I mean, is there any plan for uploading it to huggingface/datasets or publish it in any other form?

Thank you very much in advance :)

[finiteautomata]

@alexvaca0 Thanks for your interest! We will be publishing the original tweets soon, hopefully in datasets. Leave this issue open so we let you know when they are available.

[finiteautomata]

Hi @alexvaca0. I'm having some problems regarding the original tweets -- that is, the raw tweets prior to any preprocessing and filtering. The machine which contained this data is not turning on, let's hope the disk is ok.

In the meanwhile, I have access to the preprocessed and filtered tweets (as described on the paper). If that's useful for you, send me an email and I'll give access to them.

I leave this issue open until we are able to publish the original data.

[avacaondata]

Oh that would be so great, if it is still possible to have access to the tweets... thank you very much :)

[finiteautomata]

Well, this is quite late, but finally, the tweets were released. I could only upload half of them, but I suppose this might be enough (~300M tweets).

Check https://huggingface.co/datasets/pysentimiento/spanish-tweets

In the following days, I will be uploading the rest of them.