Ontology Reduction - already underway under OntoBank project (ISI)

Ontology reduction refers to cleaning up Omega

removing/conjoining some WordNet concepts
reconciling/conjoining Mikro concepts with WordNet concepts
possibly moving concepts to new locations
At ISI we have two student-level workers (one Linguistics grad student, one well-trained undergrad on vacation from Penn) working on this full-time. A few people at Penn are also doing this. (The annotators, in addition, are all at Penn.)
The methodology is as follows:
get next word (verb, noun, adj, adv)
looking at 10 to 20 example sentences, two people create the sense inventory (i.e., list of concept meanings) for that word
the adjudicator reconciles their lists
if there is a problem, a small amount of annotation is done to determine feasibility, and adjudication is finalized
we check the sense list against the ontology:
- if things are fine, we compress the extraneous WordNet / Mikro concepts
- if things are not fine, we split the wordsenses as needed by the ontology and re-adjudicate
The words are being addressed in more or less word frequency order, by part of speech, so the more polysemous words tend to come first.
The tempo is still slow -- about 20 words a week -- but is expected to pick up to several hundred words a week, ending with essentially all after 6 months (if the funding continues).
Bottom line: ontology reduction is being done more or less along the lines we'd like it for IAMTC for free!

Version 1, Fri 02 Jul 2004 14:28:21 [EH]