Re: [ontolog-forum] What goes into a Lexicon?

[Amanda Vizedom <amanda.vizedom@xxxxxxxxx>]

Many, many ontologists would say that the sense of "lexicon" that Matthew put forward is local to purely logic-oriented conversation, but as with other expressions from the many fields that contribute to applied ontology, using "lexicon" this way within the broader practice of applied ontology only creates confusion.

Rather, when we are talking about applied ontologies that include both logical and natural-language elements, "lexicon" usually refers specifically to the natural language elements. And the lexicon is most certainly *not* governed by a restriction to one lexical _expression_ per concept.  In fact, in many usages, a significant benefit of using ontologies instead of some other model type is that the structures and distinctions empowered by ontological abstraction, and by many ontology technologies, support the clear and emphatic distinction between the lexical and the conceptual. The most successful approaches, in real usages full of polysemy, make this distinction. The lexical and the conceptual are mapped; they are not the same. Understanding and implementing this distinction allows a logically precise ontology with unique concepts in which ambiguity is avoided *and* a rich representation of natural language expressions. A concept may have any number of labels (lexical mappings to words); a natural language _expression_ may be a label for any number of concepts.

Ontology projects that do not make and implement this distinction might work on a very small, or narrow-domain, or controlled environment, project. Beyond that, they will get bogged down in totally unnecessary efforts to settle on unique labels, or to choose between different concepts that share an _expression_. If they understood and leveraged the distinction, they wouldn't have to do any of that. Represent the different concepts separately, define each one appropriately (as you should anyway, as the label is not the definition, especially from a machine-usability standpoint), and map the _expression_ as a label for each of the concepts. 

There are many ways to do this, with varying levels of detail captured regarding the contexts in which expressions map to concepts. Depending on the usage, there are also varying levels of processing that can be implemented for the lexical information. For NL parsing and generation, and for text analysis and entity extraction and semantic indexing and retreival, there will typically be considerable lexical information captured and used by the NL processing engine(s). Note that this processing follows different principles, and uses different (overlapping) portions of the ontology, than does the inference engine, because *lexical information and logical information are not the same*.  For many usages, a good ontology must incorporate both kinds of information, and treat them appropriately. 

Amanda

On Sun, Feb 19, 2012 at 10:41, David Eddy <deddy@xxxxxxxxxxxxx> wrote:

Matthew -

On Feb 19, 2012, at 10:18 AM, Matthew West wrote:

I would have expected you to support them in this from your remarks, rather than be critical.

"Different perspective" would be my preferred term.

If the world of ontology would clearly & openly state: "An ontology is useful for this & not for that" I'd be far happier & less acrid.

If logic would help with the maintenance & design of software systems, I'd be all for it.  But if logic hobbles the use of terms to a single meaning, then clearly logic isn't much use in maintaining/developing systems.  The context of my view is commercial business systems, not life sciences.  I don't know if geology is closer to a life science.

As always I revert to my favorite, oft-repeated example... a life insurance company that found 70 different names for the core business concept (term?) "policy number."  In a business environment where product lines are bought & sold, systems are custom built by different teams & packages are bought, the reality is there will be many (illogical) names for the same thingy.

At one point I entertained delusions that ontologies would help with this issue (one conceptual label = many physical labels).  Obviously I no longer hope in that direction.

___________________

David Eddy

deddy@xxxxxxxxxxxxx

781-455-0949

_________________________________________________________________
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
 

_________________________________________________________________
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    (01)