Which is better? OpenCyc or ConceptNet?

Question

I'm doing a NLP project where I need to recognise concepts in sentences to find other similar concepts. I do this to infer word valences from a list I already have. I started using WordNet, but it gave many contradictory results. By contradictory results I mean word expansions that had contradictory valences.

So now I'm looking into ConceptNet and OpenCyc. I've already implemented ConceptNet and it was all very easy and I love it. Problem is that OpenCyc appears to have a much larger and more logically rigid database, which is important when I found so many "contradictions" on WordNet... But I wouldn't know because I haven't tried it.

Could someone tell me if it's worth going through the (considerable, for me) effort to implement OpenCyc, or is ConceptNet good enough to infer word valences? Are they that different?

I'll be happy to explain myself further, if needed. Trying to keep it short for now!

Thanks!

Answer 1

It's not worth the effort. ConceptNet is heaps better and easier to deal with.

When I used ConceptNet, it was a matter of downloading a sqlite database and a python module to access the relations in it. For OpenCyc, it was a bigger download, and had to be set up through a web server that took ages to load and used enormous amounts of memory. Things may have changed since, but this was my experience.