W3C

- DRAFT -

AWWSW

16 Aug 2011

See also: IRC log

Attendees

Present
DBooth, jar, +1.716.810.aaaa, alanr
Regrets
Chair
Jonathan Rees
Scribe
dbooth

Contents


<jar> http://masinter.blogspot.com/2011/08/expert-system-scalability-and-semantic.html

jar: There's stuff on the web and you want to be able to talk about it. The role of the URi is that it helps one person communicate ... helps you say what stuff you're talking about.
... If I make a statement about a document I get from a URI, you may want to know what document I was talking about, so I may want to say "the one at this URI".
... The URI is maybe part of what you need to explain what you are talking about.
... You might talk about it with a pronoun with some characteristics.

dbooth: Some metadata....

<jar> does the document contain the word "green"?

jar: At some point there will be enough info about what's being discussed to have successful communication. E.g., if I talk about a journal article you can narrow it down enough to be able to discuss further question, such as "does this article contain the word 'green'?"

<jar> might be able to identify using sha1, or author/title/date

<jar> but not adequate always for answering the "green" question

jar: Author and title may be enough to identify the doc but not adequate for discussing aspects of its content.

alan: The identification process is the step needed to acquire the doc.

<jar> do I know enough about what they're talking about to be able to answer word-containment questions?

jar: A class of strings would be defined by an invariant, e.g., same author.

<Zakim> dbooth, you wanted to ask Isn't identification the ability to know which thing is intended?

jar: But that's not useful, becuse you need an operational test.

dbooth: You want to confirm the identity enough to ensure correct communication?

jar: And i verify that by checking: if they say the document contains the word green, then I see if it is.

<jar> class of strings

<jar> satisfying some invariant

jar: lets talk about classes of strings instead of documents.
... I'm talking about some class of strings, and you don't yet know what class it is. I tell you that if it is in this class then it was written by this person and has this title.
... Or maybe it goes the other way around. I ask you something that follows from that. Here's the author-title-date . . .

<jar> game: Alice gives restrictions (invariants) for a class of strings C, she calls the class "C"

jar: You're giving property invariants for a class of strings. Alice calls the class "C" and tells Bill restrictions on members of C.

<jar> Alice tell Bill these restrictions on members of C

<jar> Alice queries Bill on some restriction, see whether he answers correctly

jar: Then Alice queries Bill on some restriction and sees whether Bill answers correctly.

<jar> Alice tells Bill, title, author, date.

<jar> Alice asks Bill, does every member of C contain the word "green"

dbooth: Does the fact that it contains "green" logically follow from the metadata that Alice gave?

jar: No, not necessarily. There has to be some other communication channel, e.g., the web, the library, background knowledge.
... There are many ways that a URI might be used in a restriction to help out like this. A URi may be the location where you GET the string.

alan: Or a GET may retrieve the hash.

jar: Or it could be a source of metadata.
... I'd like to have 10-20 different relationships between a URI and a thing.

<jar> Scenario A. Entity is a string. URI is related to it ... not sure how.

<jar> Scenario B. Entity is a class of strings. ...

<jar> TimBL case, scenario B, where URI is the one use to GET strings...

<jar> Scenario C. Entity is a source of information (HTTP endpoint)...

alan: Party 1 gives a word, and party 2 says yes or no, the class of strings contains that word

dbooth: you've got two axes: what scenario (A, B, C) and what kinds of statements are being made about it?

jar: We've looked at the special case of using a URI to help tell someone what thing you're talking about.
... The timbl case is when the URI is adequate to identify the document -- the sole piece of information you have about it.
... But there are other cases where you may need to consider time, or transclusion.

dbooth: example using only strings?

<jar> Alice says, S has SHA1 hash H. Then asks Bill, does S contain "green"?

jar: Alice says s has a sha1 hash, then asks bill does s contain "green"?

<jar> similar to

<jar> Alice says D is a generic resource timbl-identified-by URI. Asks, does D contain "green"?

dbooth: So Alice provides some characteristic about the thing, and asks if it has some other characteristic.

<jar> is being related to URI U predictive of anything?

dbooth: I think an issue here is that URIs are being used both at the meta level (as an arbitrary identifier) and at the concrete level as an attribute of something.

<jar> problem is putting the URI-relates-to-thing opportunity in perspective

alan: You've got a <URI, thing> pair, and you don't necessarily know how they're connected.
... And one party knows what thing it is i assume.
... Or knows some tru propositions about the thing, and those propositions may or may not include the URI.

<jar> Situations where, if Alice does *not* divulge a URI, then Bob willl not be competent

alan: Party A is in posession of a URI and some questions. They have a comm channel. They have some communications of mutual expected benefit.
... And if the sem web works ideally, then the first thing A would say is "the URI is u". B follows its nose. Then A asks "is having the word green a property of it". and B says yes. And they convince themselves that they've been successful in communicating.
... Another scenario: A tells B "you know that person who has red hair and lives at so-and-so?" Here's a URI, and give it to another person, and that person gives you a URI of the thing.

dbooth: How does that convince A and B that the've communicated successfully?

<jar> (competence test)

alan: That second URI allows them to answer any questions about the thing.

<jar> flickr case: the relationship between the landing page URI and the image

<jar> jamendo: the relationship between the landing page URI and the music

<jar> jamendo says, the music "with" URI U, has license L

<jar> "with" relationship is .... ?

Summary of Action Items

[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.136 (CVS log)
$Date: 2011/08/16 18:32:55 $

Scribe.perl diagnostic output

[Delete this section before finalizing the minutes.]
This is scribe.perl Revision: 1.136  of Date: 2011/05/12 12:01:43  
Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/

Guessing input format: RRSAgent_Text_Format (score 1.00)

Succeeded: s/benefit/mutual expected benefit/
No ScribeNick specified.  Guessing ScribeNick: dbooth
Inferring Scribes: dbooth

WARNING: No "Topic:" lines found.

Default Present: DBooth, jar, +1.716.810.aaaa, alanr
Present: DBooth jar +1.716.810.aaaa alanr
Got date from IRC log name: 16 Aug 2011
Guessing minutes URL: http://www.w3.org/2011/08/16-awwsw-minutes.html
People with action items: 

WARNING: No "Topic: ..." lines found!  
Resulting HTML may have an empty (invalid) <ol>...</ol>.

Explanation: "Topic: ..." lines are used to indicate the start of 
new discussion topics or agenda items, such as:
<dbooth> Topic: Review of Amy's report


[End of scribe.perl diagnostic output]