Inconsistent field names: `obj_id`/`sub_id` in docs, but `obj_uri`/`sub_uri` in dataset
Hi, thank you for this great multilingual LAMA dataset!
I noticed a discrepancy between the documented field names and the actual dataset structure.
According to the dataset card, each instance has the following fields:
"obj_id"(object ID)"sub_id"(subject ID)
However, when I load the dataset, these columns do not exist. Instead, I see:
"obj_uri""sub_uri"
Here’s the column output from the loaded DataFrame:
['uuid', 'lineid', 'obj_uri', 'obj_label', 'sub_uri', 'sub_label', 'template', 'language', 'predicate_id']
Could you clarify whether:
- This is simply a mismatch in documentation?
obj_uriandsub_uriare indeed equivalent toobj_idandsub_id?
Clarifying this would help ensure correct referencing when constructing unique fact identifiers (fact_id) across languages. Thanks again!
ping @pdufter .
Good catch
@muhammadravi251001
This is a typo in the dataset_card. It should simply be *_uri instead of *_id.
Alright, thanks for the clarification
@pdufter
! With this information I can be sure that I can create the fact_id using this f-string: f"{sub_uri}_{predicate_id}_{obj_uri}".
Can you make a PR so i can merge?
Or do I need to add that fact_id also on the dataset? I'll do that if you say so.