Open caufieldjh opened 2 years ago
I've been thinking about this and I'm not certain it would confer any graph ML benefit, at least not without specifically using antonyms as a negative training set, and that option still doesn't seem right because the terms are still closely related.
Describe the desired behavior
Ingested edges with the "is opposite of" annotation (http://purl.obolibrary.org/obo/RO_0002604) should be included while retaining this property, though at the moment this is not handled distinctly from any other predicate. It would be great to have opposites, though!
Additional context
See the current Ubergraph approach here: https://github.com/INCATools/ubergraph/blob/f078fd8969bff875127efd118783324ae631c6e4/Makefile#L60-L62