Entity list and relation list

[kaihuchen]

It seems that two files needed by train.py and test.py (i.e., through the --ent and --rel arguments), i.e. the entity list and the relation list, are not included in this repository.

Can I assume that:

• The relation list should contain the list of known relation names (18 of them in the WordNet samples, e.g., _member_of_domain_topic, _also_see, etc.)?
• The entity list should contain the list of known entity IDs (e.g., 14854262, 00590383, etc.) which can be extracted from, say, wordnet-mlj12-train.txt?

It is easy enough for me to write my own code to generate these files, but I just want to make sure that the above are what's expected. Thanks!

[mana-ysh]

Hi @kaihuchen

Your assumption is right. Entity/Relation list files contain each ID or names line by line and are generated from training triplet file. (If you run and trace preprocess.sh, it is more easy to understand, I think.)

[kaihuchen]

@mana-ysh

Manabe San, much appreciated for the confirmation!