On 3/16/12 1:29 PM, John F. Sowa wrote:
> Kingsley and Doug,
>
> KI
>>> ... and the whole lot can be mapped to URIs.
> DF
>> This is trivially true. Each program module which defines terms could
>> be given a unique namespace (NS) and then each term (T) defined in that
>> namespace could be assigned the URI: NS#V .
> Yes. That can be done with every logic and ontology since Aristotle.
>
> KI
>> Today, we hear a lot about BigData or 'Big Data' and very little about
>> the fundamental realities associated with:
>> 1. exponential growth of data volume
>> 2. exponential growth of data velocity
>> 3. exponential growth of data variety (heterogeneity)
>> 4. exponential growth of data location disparity.
>>
>> You can't deal with these matters without URIs in your arsenal.
> As I said many times, URIs are useful. But I also emphasize that the
> actual *words* that people use in any NL text are usually more *precise*
> and more *informative* than any URIs that some annotator (human or
> computer) has assigned to them. (01)
I don't dispute that. Re. Linked Data we integrate the contributions of
many NLP services [1] these days. Thus, as part of a Linked Data
generation pipeline NLP services contribute to final graph enhancement.
These services sometime contribute URIs in addition to basic entity
extraction. (02)
>
> Even for annotators who are professional linguists and lexicographers
> 95% agreement is considered exceptionally good. Typical data entry
> clerks are so clueless that their annotations are worthless.
>
> The methods for processing NL texts are improving rapidly. Even
> today, the best NLP systems can do as well or better in selecting
> word senses than the average human annotator. (03)
Yes, as per my comments above re. taking this effort and meshing it with
Linked Data. (04)
>
> Please note: I do *not* claim that current NLP systems are good
> at selecting word senses. I am just observing that typical human
> annotators are horrible. (05)
Understood. (06)
>
> If anybody thinks that there is a contradiction between my faith in
> human authors and my skepticism about human annotators, I'd like to
> emphasize a fundamental principle that linguists have observed:
>
> People who have never studied linguistics or logic can use their
> native language correctly when they don't think about it. But
> when somebody asks them a question about grammar or meaning,
> their answers are hopelessly confused.
>
> Conclusion: Don't depend on human annotators to select URIs. (07)
Amen! (08)
> And if you are processing NL texts, it's better to have your
> system select word senses than to depend on any annotations
> in the texts. At least you can control your own system, but
> you have no idea where any annotations may have come from. (09)
Links: (010)
1. http://linkeddata.uriburner.com/c/IHWRTT -- note the 'mentions'
property values which are qualified by NLP service contributor (011)
2. http://linkeddata.uriburner.com/c/IH22OE -- alternative faceted
browsing and follow-your-nose navigation friendly view (012)
3. http://ode.openlinksw.com/ - browser extension and/or bookmarklet
that I use to produce these Linked Data Pages (013)
4. http://ode.openlinksw.com/example.html -- usage example (014)
5. http://uriburner.com -- public version of Linked Data middleware
service used by the above. (015)
Kingsley
>
> John
>
> _________________________________________________________________
> 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
>
> (016)
-- (017)
Regards, (018)
Kingsley Idehen
Founder& CEO
OpenLink Software
Company Web: http://www.openlinksw.com
Personal Weblog: http://www.openlinksw.com/blog/~kidehen
Twitter/Identi.ca handle: @kidehen
Google+ Profile: https://plus.google.com/112399767740508618350/about
LinkedIn Profile: http://www.linkedin.com/in/kidehen (019)
smime.p7s
Description: S/MIME Cryptographic Signature
_________________________________________________________________
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)
|