Re: Troponym Structures?
Last week I queried the SUO group and the CG list about inheritance
structures built around verbs. A number of people have sent me
suggestions for such ontologies. These include existing tools (VerbNet,
et. al.), current research, and old standbys like Bloom's taxonomy. This
has been helpful; thank you all.
Some, especially Arun Majumdar, noted that ontologies are also built
around meronyms, entailments, and so on. I am aware of that. I'm also
dimly aware of at least some of the problems listed by Avril Styrman
when focusing on the predicates, rather than the arguments of a
relation. John Bateman observed that clauses, rather than verbs, might
be more useful. Clauses may be especially appropriate for my purposes,
as will be shown in the following description of my problem.
The reason I wish to focus on troponyms is simple: first things first.
Below is a general description of the situation I am working through,
and why it leads me to examine verbs in formal structures.
System engineering problems I encounter often are caused by bad
precision in requirements declaration, or poor translation into
specifications. The problem domains here range from hardware systems and
software systems (or both), management processes viewed as systems, and
systems of subsystems. The manifestation is degradation or failure,
usually at the interfaces. The interfaces in some cases were unexpected.
Usually, a lot of information is available from specifications, emails,
problem reports, meeting minutes, and all the detritus a project throws
off. I generate concordances to select key words, often several dozen of
them. The collocations of the words highlights similar usages, synonyms,
euphemisms, and so on, which must be resolved. The results of the
concordance are then graphed as simple semantic networks. The semantic
nets are aids in finding hidden associations and consequences. As often
as not, how the words are used suggests a focus on the verbs, that is,
the predicates, as a structuring mechanism.
I want to add additional structure to the simple semantic nets. Since
most of the words are IsA types, I decided first to focus on troponyms.
PartOf relations would be the next to consider, but as I noted above,
first things first.
Regards,
-- Jeff Schiffel