Azamat, (01)
I agree with your criticisms of the Cyc upper ontology: (02)
AA> Doug's dinner is hardly an ontology cooking of haute cuisine.
> Rather, it is a concoction of disagreeable ingredients without
> a base of stock. The whole stuff is made without a conceptual
> base, foundation ontology. (03)
I made very similar criticisms in the early 1990s, and the Cyc
upper levels have not improved much, if at all. (04)
AA> ... there are two very simple criteria to test the real value
> of schemes, knowledge systems, languages, models, or ontologies;
> namely, see how:
>
> 1. they treat the category of Thing or Entity;
> 2. they define the class of Relationships. (05)
I agree that those are two important criteria. But I would add
two more: (06)
1. Has it been implemented and used successfully on a
significant number of practical applications? (07)
2. Is the upper ontology necessary (or at least useful)
for the practical applications? (08)
Cyc has been under development and use for 24 years, and it
probably has had more practical use in more ways than any
other large formal ontology. The WordNet system (which we
can all agree is not a formal ontology) has had even more
use in many more practical applications. (09)
But Lenat himself has said that the upper levels were
relatively unimportant compared to the middle and lower
levels. Nobody has had a kind word to say about the upper
levels of WordNet, but the middle and lower levels are
very widely used. (010)
AA> ... the CYC failure shows one plain thing: a strong need
> for standard or foundation ontology, to head off large-scale
> knowledge engineering projects. (011)
You haven't given a single reason for that claim. Note that
Cyc and WordNet, despite their poor upper levels, have been
used for a large number of practical projects -- far more
than any ontology with a better (or at least different)
upper level. (012)
My recommendations: (013)
1. Guidelines and tutorials about how to define an ontology
for specific low-level domains. (014)
2. Lexical resources such as WordNet, Roget's _Thesaurus_,
and others for relating the words of natural languages
to the categories of the mid-level and low-level domains. (015)
3. A large collection of modules (such as the Cyc microtheories)
for various purposes. (016)
4. Metalevel relations among the modules (such as generalization,
specialization, and other links). (017)
I wouldn't claim that upper levels are totally irrelevant, but
that there is no unique upper level that would be ideal for all
purposes. Rather than try to improve the Cyc and WordNet upper
levels, I would ignore them. (018)
John (019)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx (020)
|