- From: Butler, Mark <Mark_Butler@hplb.hpl.hp.com>
- Date: Thu, 11 Dec 2003 17:13:04 -0000
- To: "'SIMILE public list'" <www-rdf-dspace@w3.org>
--> You are now talking on #simile --> SteveGarland (garland@18.52.1.28) has joined #simile <marbut> attendees: eric, steve, mark <marbut> mark: any progress on getty? <marbut> ericm: no, no progress <marbut> mark: Mick was keen for us to obtain a copy <marbut> ericm: I'll put that on my to do list <marbut> SteveGarland: as I mentioned last week, 2 students working on things <marbut> Vineet's faceted browser can show artstor data in Haystack <marbut> he's going to be available for a bit of time on Wednesday this morning, but he's supervising an exam, so not available for all 2 days <marbut> David H thought it would take a short time, to link Haystack to Joseki, but only with help from Denis, <marbut> but after some email exchange, it looks like it easier than he thought <marbut> also I've made some progress on a web browser front end for the artstor data <marbut> a couple of technical issues still need to be resolved <marbut> what to get with ungettable URIs, how to display them --> ksmathers (chatzilla@206.170.148.113) has joined #simile <marbut> also some issue about dublin core versions, theres a slight change to from version 1 to version 1.1 <marbut> I'm not familiar with the details, the title changes in DC in the ontology, I think Haystack is using 1.0, so its <marbut> not finding things coded in the database with 1.1 <marbut> ericm: Haystack should be using 1.1 <marbut> The 1.0 / 1.1 problem is fairly common. 1.1 came into being three years ago. You can introduce subProperty / owlEquiv to jumpstart things. <marbut> Mark: Once we have the Joseki link, we can do that at the Joseki end <marbut> SteveGarland: Once David H gets his bit done, that's not a problem <marbut> ericm: Can we add ungetable URIs to the agenda --> RRSAgent (rrs-loggee@18.29.0.30) has joined #simile * RRSAgent is logging <ksmathers> mark: minor changes to artstor; uri's for instances all point to locations on simile web site. put up some minimal pages <ksmathers> mark: indicating that the namespace is reserved. <ksmathers> eric: any other plans for namespace. <ksmathers> mark: uri's are used for anything that is likely to be crossmapped. contributors, subjects, keywords <ksmathers> mark: identified commonality in the data, then an owl equiv can mapbetween the common instances in the data. <ksmathers> mark: also worked on the demo script. kevin wrote a longer script scenario and events. mark has done two more in the same style. <ksmathers> mark: also working the team meeting in boston <marbut> kevin: I've been working on the 20th century art course <marbut> this has a much stronger overlap with artstor than the other courses we have <marbut> it more than doubles the number of resources, and might be ten times as many overlaps <marbut> unfortunately it didn't have any resources defined only class nodes <marbut> because OCW couldn't obtain rights to these resources <marbut> so I took the lecture notes, took the information out, and did a hand mapping to see what the overlap would look like <marbut> I put this in an excel spreadsheet, then I've converted it to OCW LOM XML, for each of the items in the spreadsheet <marbut> so it now has 500 new resources <marbut> this is not checked into the repository <marbut> they should map well over to Artstor, there are much stronger overlap with the others <marbut> there are some limitations on the mapping, that I've described in my email, the descriptions are weak, but if we get a strong mapping to artstor <marbut> we can copy them from artstor. I don't think its worth doing it for all 500 images <marbut> the second limitation is because OCW didn't have any location information, I did websearches for the technical information <marbut> I tried to locate instances of these images on the web. So a lot of the URLs are on the public web, we might want to copy them down, <marbut> in case they go away. <marbut> Mark: What are the copyright on the images? <marbut> mark: let's talk about ungetable URIS <marbut> mark: eric, we'd be interested to hear your position on this? <marbut> ericm: so does haystack need objects of a particular kind, or do they just need to avoid 404's? <marbut> SteveGarland: not sure, I think they just need retrievable objects <marbut> ericm: but if they got more RDF, could they deal with it? <marbut> SteveGarland: yes, it can deal with it <marbut> ericm: the current plan is we are sticking a HTML page? <marbut> kevin: these are RDF objects, that have URLs associated with them, they can point to the objects themselves, we could <marbut> just use Joseki to give you back the objects themselves <marbut> ericm: is this debatable? <marbut> kevin: I'm not sure if Andy likes it <marbut> He says its not a publishing system, its a query system <marbut> He doesn't want to deliver large bits of RDF from Joseki <marbut> but I think you can just deliver a localised region, a data object <marbut> in joseki, there is a concept of a fetch <marbut> ericm: but it also strikes me that we could have the objetcs <marbut> outside of the server, on a normal server <marbut> kevin: but then the information might get out of sync? <marbut> ericm: Yes but any sort of web harvesting might do that. If you do decouple it, that's the penalty we pay, <marbut> Kevin: I'd like to see one canonical source, and people get data from the URL e.g. the Joseki server <marbut> ericm: So we had a similar architecture in the Pearl stuff <marbut> Kevin: we can talk about this next week <marbut> ericm: That's a good idea, what you say makes sense, we haven't talked about the notion of RDF harversters yet, but in the future <marbut> people may be interested in integrating our data with other sources a lot easier <marbut> ericm: I think there are only three solutions here, so there are not too many solutions <marbut> mark: at least if we have discussion in the morning, he will be on the telecon <marbut> ericm: Has this discussion moved to controlled vocabularies in this context? <marbut> Kevin: Mark and I had a long discussion off list about whether keywords should be placed in URI space <marbut> So now for expediency we are turning them into URLs, but I think there is a mismatch between URLs and keywords <marbut> ericm: It depends, some people think of keywords in controlled vocabulary sense <marbut> Kevin: I do think of them in that way, but as we are not an authority I don't think pointing them at our URL is useful <marbut> It's easy to find utilties to do things like word stemming, case folding, the hierarchies etc <marbut> ericm: But they are not mutually exclusive <marbut> Kevin: That's why we tabled it, <marbut> ericm: We need labels on the keywords <marbut> Kevin: I need to add labels - which namespace? <marbut> Mark: RDFS <marbut> ericm: We don't want to lose information here. <marbut> That makes a really compelling demo, when the librarian can set up equivalence relationships, it turns out to be a powerful way of leveraging both free text tools and semantic web tools <marbut> Kevin: Should I put a language indicator in? <marbut> ericm: Yes, if its right. Adding language there is a good thing. This will be useful in phase 2 when we connect with CIDOC <marbut> Mark: They get it wrong in CIDOC - all the latin and german is marked English <marbut> ericm: Yes, but in the future we need to get it right <marbut> Does CIDOC get it wrong? <marbut> mark: yes it just says english at the top, but it uses a mixture of german and english <marbut> http://web.mit.edu/simile/www/documents/demoscript/demoscript.htm <marbut> Kevin: script 1 - I need someone who is both a learner and a teacher <marbut> I need someone with access to the Artstor db, as they dont share their data with everyone <marbut> hence the actors <marbut> the scenario deals with a course on art history <marbut> 20th century art is relevant to things like advertising, so I think this is relevant to a two year college <marbut> SteveGarland: I really like the background in scene 1 here. Some of the concerns may be state specific, but I don't think that matters <marbut> I found it really helpful on focussing on who the users are going to be. <marbut> SteveGarland: The problem you raise is a real one. <marbut> ericm: These kind of demos apply social / peer pressure to make that a reality. <marbut> kevin: because this is the mode of artstor, all of the resources artstor has are owned by one of the museums / colleges in the consortium <marbut> ericm: it a bit more complicated, but let's stay with that <marbut> ericm: I hope that people will realise thumbnails are metadata <marbut> mark: can we hear steve questions? <marbut> SteveGarland: they are more about how does this work? <marbut> this guy visits his friend, and is then able to access information he couldn't before, how does that work? <marbut> Kevin: Its just he didn't know, he knew about the data before, but not how to tie them together <marbut> SteveGarland: So the situation he was in that if he treated them as separate databases he would have a lot of extra work. <marbut> There's lots of stuff you kind find, but how does it compare with the amount of work he would need to do before? <marbut> I have the same question about all three scripts. People can do this kind of thing already, with Google, the question is how is it improved with SIMILE <marbut> ericm: I think this is a good idea, we need to know how these things are better than before <marbut> or if one person makes the connections, we need to record it so the next person doesn't need to do the work <marbut> kevin: This was all background originally, so I was going to fill in with actual steps that the actors took <marbut> SteveGarland: That would be great <marbut> ericm: BTW, I really like this direction <marbut> If we incorporate the value added dimension, then we have strong justification for the demos we are doing <marbut> We need to ground the SW in real applications <marbut> Kevin: There was one thing that Mark brought up, in that it starts with OCW and goes to Artsot <marbut> that's because its easy to justify <marbut> It's harder to go the other way <marbut> ericm: What courses are this image part of? <marbut> Art professors are like curators, so to the extent other people have done this, that's value added for them <marbut> but you can easy to see that connection there <marbut> SteveGarland: its just as if we were teaching a course of computer science? <marbut> kevin: But why go to artstor to begin with, not IMS? <marbut> ericm: Start with Artstor, then ask the question: what's the most popular resources in courses? <marbut> It's a kind of link analysis <marbut> SteveGarland: I have an image, then I want to know how different organisations deal with it in their curriculum. <marbut> ericm: i don't think we need to flush out all the scenarios to a high level detail, but I'd suggest you put in question marks, then throw it out to the rest of the group <marbut> SteveGarland: For example in OCW, you might have two professors teaching the same course, I want to compare the syllabi <marbut> There are alsorts of useful things that can be done here. Dr Mark H. Butler Research Scientist HP Labs Bristol mark-h_butler@hp.com Internet: http://www-uk.hpl.hp.com/people/marbut/
Received on Thursday, 11 December 2003 12:13:42 UTC