nmrksic / attract-repel

The Attract-Repel algorithm presented in (Mrkšić et al., TACL 2017), with accompanying resources.
Apache License 2.0
64 stars 17 forks source link

Maybe misuse of the spearmanr? #4

Open ryh95 opened 4 years ago

ryh95 commented 4 years ago

Dear authors of AR

I noticed that you first rank the list and then use the scipy.stats.spearmanr in your evaluation function simlex_analysis.

However, someone directly uses the spearmanr without ranking, according to https://stackoverflow.com/a/14819086/6609622 and https://machinelearningmastery.com/how-to-calculate-nonparametric-rank-correlation-in-python/

Basically, they said the spearmanr will take care of the ranking for you.

So, maybe the spearmanr here is misused?

Thank you!