WebBrain jointly creates representations of words and relations using neural methods. It does this based on massive amounts of text, exploiting both general co-occurrence patterns, and much more specific patterns that reveal relationships between words or word meanings. For instance, the fact that grass is often green, or that water is something that we may drink.
For more information about our data, please proceed to the data page. Additionally, consider consulting the WebChild project data as well.
For more information about the method, please consult our publication:
WebBrain: Joint Neural Learning of Large-Scale Commonsense Knowledge PDF BibTeX
Jiaqiang Chen, Niket Tandon, Charles Darwis Hariman, Gerard de Melo (2016)
In: Proc. ISWC 2016.
Acceptance rate: 18%