Given an entity (e.g. of type PERSON) with an associated relational schema (a set of relations, e.g. city-of-birth(PERSON, CITY), schools-attended(PERSON, ORGANIZATION), spouse(PERSON, PERSON)), all relations about the entity that are expressed in a text corpus would be relevant, and the correct answers would have to be extracted.
Currently, RelationFactory comes with pre-trained models for the schema of 41 relations defined in the TAC KBP Slot-filling task.
The source distribution can be downloaded at: https://github.com/beroth/relationfactory/