malllabiisc / RESIDE

EMNLP 2018: RESIDE: Improving Distantly-Supervised Neural Relation Extraction using Side Information

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

How do you get type_info.json?

YaNjIeE opened this issue · comments

Hi, guys:
Thanks for the contributions.
Now I'm a little confused about how do you get the type_info.json.
Could you please tell me about it?

Thanks a lot.

Hi @yanjieg,
I got the entity type information from Freebase data dump and then remapped it using FIGER.

Awesome! Thank you for reply.
there are another two questions I want to ask:
①I notice there are totally 87 types in your type_info.json, but the FIGER has 112 (or 113?) types in total. There are still 25 types out there, does this mean that the NYT entities don't include this 25 types?
②Would you tell me how you remap it ? I notice that there are some type information about entities in NYT10 dataset, for example, the entity 'Italy' has type '/location/dated_location', '/aviation/aircraft_owner' , '/user/robert/military/topic', '/location/country' and so on, and the FIGER also has '/location/country' type, so we can just say that 'Italy' has the type '/location/country'. So, you remapped it by comparing the types in NYT10 dataset and FIGER based on strings?

Thank you and look forward to your reply.

Hi,

  1. As far as I remember, I manually merged a few of the types as they were very similar to each other.
  2. In RESIDE, we allow each entity to have multiple types. We use an average of embeddings corresponding to the types it belongs. FIGER provides a mapping from types provides by Freebase to its 113 categories. If 'italy' has multiple entity types from Freebase then we just remap all of them using mapping from FIGER.

Please let me know if you need any further clarification.

Thanks

Hi, @
It's very nice of you, thank you very much.

Have a good day : p

No problem.
Same to you :)