Hello!
I am currently trying to reproduce the results described in "Inferring psycholinguistic properties of words" (https://www.aclweb.org/anthology/N16-1050.pdf), by implementing the authors' bootstrapping algorithm. However, for some unknown reason I keep getting correlations with the actual ratings in the .4-.6 range, rather than the .8-.9 range. If you have a software implementation that reaches the performance levels from the paper, could you please share it with me?
Many thanks,
Armand