GOLEM-lab / fandom-wiki

Extraction of structured and unstructured information from fandom.com pages
10 stars 0 forks source link

Implement object extraction filtering #5

Open txetxedeletxe opened 1 year ago

txetxedeletxe commented 1 year ago

The object extraction QA system may often extract texts that are not valid as an object of the target relation. These can easily be filtered by imposing static restrictions on the types of object that can be extracted in terms of their PoS and, when an entity, the properties of this entity.

This issue depends on: https://github.com/GOLEM-lab/fandom-wiki/issues/3 https://github.com/GOLEM-lab/fandom-wiki/issues/4

For now implement basic filtering heuristics. More elaborate criteria may be implemented when the KB is in place.