How does entity-fishing manage to assign a "Domain" to an entity ?

[aa303554]

In the example, Switzerland has the domains: Geology, Oceanography, Earth. Does this domain correspond to a wikipedia category or a wikidata category or are the domains obtained in another way ?

[kermitt2]

Hello @aa303554 !

The domain is assigned based on a mapping using the high level category hierarchy of the English Wikipedia. If I remember well the mapping is there -> https://github.com/kermitt2/entity-fishing/blob/master/data/wikipedia/mapping.txt

The motivation is that we have around 1 million categories in the English Wikipedia and it's very hard to get a clue about the domains. Domain information is also found in many ontologies and it can help mapping between knowledge base.

Using the domain could be also a feature interesting to use for the disambiguation by the decision tree, but it turned out that no, it doesn't help.