dotnet / dotnet/machinelearning-samples
wikiDetoxAnnotated40kRows.tsv unreliable?
Open
need-further-analysis
- Dominant language
- PowerShell
- Stars
- 4.7k
- Forks
- 2.7k
- Avg merge
- 2d 22h
- Merged PRs (30d)
- 1
Description
Not sure where to post this. but the wikiDetox dataset used in the tutorial for Sentiment Analysis seems unreliable. 22k neutral vs 2k negative comments gives a very low recall. Having an function that would just returns false for every testing data would have a precision of 90%.
Contributor guide
Assessment
This issue has not been assessed yet.