Input files for the code

[soummyaah]

Hi,

I am extremely interested in reproducing the code from your paper. However, the initial datasets found at JAPE only contain 5 files: ent_ILLS, att_triples for each language and rel_triples for each language. Can you explain the process you use to get training_attrs files?

[1049451037]

It is extracted from the initial dataset from http://ws.nju.edu.cn/jape/

[soummyaah]

Yes, I understand. Can you explain the extraction method? Particularly for the training_attrs files.

[1049451037]

Oh, sorry. I find the training_attrs is the original data file from https://github.com/nju-websoft/JAPE/blob/master/data/dbp15k.tar.gz

I didn't modify it.