On 20News data processing

ZixuanKe / PyContinual

PyContinual (An Easy and Extendible Framework for Continual Learning)

300 stars 62 forks source link

On 20News data processing #5

Closed Chen-Hailin closed 2 years ago

Chen-Hailin commented 2 years ago

When I re-run the existing system for 20news, I find that the input data text is like: "Newsgroups: rec.motorcycles\nPath: cantaloupe.srv.cs.cmu.edu!ro ...etc".

Am I right that you do not discard the header (which often contains the name of the newgroup label) during data processing?

ZixuanKe commented 2 years ago

Hi Hailin,

Thank you for your interest! Yes. We didn’t apply any pre-processing to the 20 newsgroup data.

Feel free to re-open if you have further questions.

Thank you for your time, Zixuan