I'm looking for a dataset which will give me is-a and has-a type relations between senses (and give senses from words). WordNet seems to be the most popular model, but I'm trying to evaluate other possible sources (also WordNet doesn't support slang, txt-speak, or company/brand names, which I will eventually need, either in the database I use or merged into it somehow).

To clarify, I'm looking for data which I can process to give relations roughly like:

Chair: is-a Seat is-a Furniture is-a Physical Object is-a Entity

Finger: (part-of Hand part-of Human is-a Animal is-a Physical Object is-a Entity) OR (is-a Measurement Unit is-a Abstract Entity is-a Entity)

I haven't been able to find anything beyond WordNet or dictionary-scraping so far. Any suggestions?

  • What exactly do you mean with "sense"? I don't see any "senses" in Chair: is-a Seat is-a Furniture is-a Physical Object is-a Entity. – Christian Oct 19 '13 at 13:04
  • @Christian, I mean like the second example, where "finger" can be in the sense of a part of a hand, or in the sense of a unit of measurement (silly example but you get the idea), but once you pick one and start going through its hierarchy, you always know which sense of each node word you're talking about. If that makes sense! – Dave Oct 19 '13 at 15:20

ConceptNet is a semantic network containing lots of things computers should know about the world, especially when understanding text written by people.

Trying to reproduce your relation-sequences yields:

| improve this answer | |
  • Thanks. It seems a bit too unstructured though; is there a reliable way to navigate it? From my quick tests, it doesn't distinguish senses of words and has no obvious path towards more general things (the links can be more general, more specific, or alternative forms, with seemingly no distinction between the types). Did I miss something? – Dave Oct 19 '13 at 12:29
  • I fear that is by (lack of) design. I consider the database "research grade", i.e. work in progress. – ojdo Oct 21 '13 at 9:11

English EuroWordNet at globalwordnet.org is worth considering, depending on your requirements.
Opencyc.org and Google freebase.com have large taxonomies already available for download.

| improve this answer | |
  • Please add links as links – ojdo Oct 21 '13 at 9:10
  • OpenCyc looks very interesting; I'm trying to find an OWL viewer which can cope with all its data! – Dave Oct 21 '13 at 10:39
  • 1
    OpenCyc has been killed off. – chicks Sep 14 '17 at 19:48

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.