First attempt at a cleaning

#2
by mike-ravkine - opened

Cleaner code: https://gist.github.com/the-crypt-keeper/94e72696ff4f2ba0355e6bfc83e97341

560 of the input articles were converted into 1284 rows of {title, language, task, solution}

very much a work in progress, dont use yet :)

Weird HF glitch caused a double-post, lets use this one.

Stand-alone dataset uploaded here so it can be explored: https://huggingface.co/datasets/mike-ravkine/rosettacode-parsed

Lots of low hanging fruit still remaining in terms of cleaning, not to mention the fact that only ~10% of the articles seem to be in the right format (need to look into that..)

I didn't actually look very closely at the raw scrape data yet, could be a bug in what I did. Some pages aren't actually tasks either, haven't had a chance to filter those out yet.

This comment has been hidden
jondurbin changed pull request status to merged

Sign up or log in to comment