DanC: a playground for SPARQL demo services and related events
danbri_: As far as I can tell, this is being presented as a profile of a spec that is as-yet unpublished/unwritten. It's a draft of a language we could use for annotating media files (eg. spoken word) with subtitles etc.
crschmidt: In looking at RAP, it seems that it doesn't have a "download model from URI" function.
crschmidt: Although the tool that I use (Redland) offers this support, I didn't end up using it for reasons I can't remember at the moment.
crschmidt: Which RDF tools do you use? Do they have Download support? If so, do you use it or something else like curl?
crschmidt: Talliesin did a survey on this on one mailing list or another recently: Do your tools use Accept: headers to specify a preference for application/rdf+xml ?
Talliesin: FOAF Explorer was the only tool reported as not specifying application/rdf+xml, but I think I may need to ask again actively seeking tools which fail to do this.
mortenf: FE does now (application/rdf+xml, /)
dajobe: redland/raptor/rasqal all prefer application/rdf+xml (higher q) but then accept others at lower q
crschmidt: julie does the same as FE (application/rdf+xml, /)
Talliesin: Main thing is does it have application/rdf+xml at a higher q than /, text/html, application/xhtml+xml, text/xml, application/xml (unless it's a tool which has some good behaviour for them, and for RDF - i.e. it isn't just an RDF tool and it won't be "upset" if it doesn't get RDF)
crschmidt: Although the tool that I use (Redland) offers this support, I didn't end up using it for reasons I can't remember at the moment.
crschmidt: Which RDF tools do you use? Do they have Download support? If so, do you use it or something else like curl?
crschmidt: Talliesin did a survey on this on one mailing list or another recently: Do your tools use Accept: headers to specify a preference for application/rdf+xml ?
Talliesin: FOAF Explorer was the only tool reported as not specifying application/rdf+xml, but I think I may need to ask again actively seeking tools which fail to do this.
mortenf: FE does now (application/rdf+xml, /)
dajobe: redland/raptor/rasqal all prefer application/rdf+xml (higher q) but then accept others at lower q
crschmidt: julie does the same as FE (application/rdf+xml, /)
Talliesin: Main thing is does it have application/rdf+xml at a higher q than /, text/html, application/xhtml+xml, text/xml, application/xml (unless it's a tool which has some good behaviour for them, and for RDF - i.e. it isn't just an RDF tool and it won't be "upset" if it doesn't get RDF)
bblfish: returns a very nice html presentation of my OWL file
libby: a preprint of the paper Gunnar's giving now at ISWC 2004 in the sw mining track (found by Simon Price - thanks Simon!)
libby: "SW topologies: SW forests - xml will do; semantic webs are different, interconnected graphs with no clear top level node - like foaf"
libby: "foaf 6.5 m triples; 259298 known foaf files as of sept 2004; many hand generated but e.g. ecademy, livejournal also producing many files"
libby: "foaf is scruffy - human errors, interest is a problem - some people just use a string and people use different urls for the same concept"
libby: cleaned it up. wanted to find groiups of people; then concert to prolog and use with ILP. then want to feed it through again with generated rules but not done that yet"
libby: "use hierarchical aggolomerative cluster algorithm (HAC), building a tree by recursively merging most similar clusters"
libby: "need similarity measures for RDF grphs. "hamming distance" didn;t work very well (immediate properties only); tried conceptual graph comparision - Montes-y-Gomez 2000; must extract subgraph around the person to use it"
libby: "trial and error - extract subgraph by 2 steps forward and one step backwards (any property) - works quite well for foaf; other data (e.g. imdb) didn't work with this heuristic"
libby: "used aleph ILP system. Uses 100 most frequent preds but not 'knows' - or else clusters are all based just on knows. the rules generateed were also constrained by the foaf schema"
libby: "rules found: (1) people who have written a particular paper together; (2) people who have group homepage www.aktors.org (3) people with nearest airport ABZ (i.e. aberdeen agents group) (4) people who trusthighly someone else (only used by 8 people though :) (5) people who created a paper of type application/postscript"
libby: "filtering of pred/value pairs might be a good idea. foaf is too sparse so get weird results, same preds are not used enough. now trying IMDB which is much denser. trying to make a movie-trivia learner - e.g. did you know that all the films you like have someone in them who dated julia roberts? use common sense and ontologies to filter predicates"
libby: "conclusions: lots of preprocessing required, scruffiness is a problem. evaluation of rules is difficult - end up with lots of them but few are interesting. scaling up is the biggest problem of the sw - some of these experiments took weeks..."
libby: "SW topologies: SW forests - xml will do; semantic webs are different, interconnected graphs with no clear top level node - like foaf"
libby: "foaf 6.5 m triples; 259298 known foaf files as of sept 2004; many hand generated but e.g. ecademy, livejournal also producing many files"
libby: "foaf is scruffy - human errors, interest is a problem - some people just use a string and people use different urls for the same concept"
libby: cleaned it up. wanted to find groiups of people; then concert to prolog and use with ILP. then want to feed it through again with generated rules but not done that yet"
libby: "use hierarchical aggolomerative cluster algorithm (HAC), building a tree by recursively merging most similar clusters"
libby: "need similarity measures for RDF grphs. "hamming distance" didn;t work very well (immediate properties only); tried conceptual graph comparision - Montes-y-Gomez 2000; must extract subgraph around the person to use it"
libby: "trial and error - extract subgraph by 2 steps forward and one step backwards (any property) - works quite well for foaf; other data (e.g. imdb) didn't work with this heuristic"
libby: "used aleph ILP system. Uses 100 most frequent preds but not 'knows' - or else clusters are all based just on knows. the rules generateed were also constrained by the foaf schema"
libby: "rules found: (1) people who have written a particular paper together; (2) people who have group homepage www.aktors.org (3) people with nearest airport ABZ (i.e. aberdeen agents group) (4) people who trusthighly someone else (only used by 8 people though :) (5) people who created a paper of type application/postscript"
libby: "filtering of pred/value pairs might be a good idea. foaf is too sparse so get weird results, same preds are not used enough. now trying IMDB which is much denser. trying to make a movie-trivia learner - e.g. did you know that all the films you like have someone in them who dated julia roberts? use common sense and ontologies to filter predicates"
libby: "conclusions: lots of preprocessing required, scruffiness is a problem. evaluation of rules is difficult - end up with lots of them but few are interesting. scaling up is the biggest problem of the sw - some of these experiments took weeks..."
libby: by Aleksander Pivk, Philipp Cimiano, York Sure
libby: paper presented at ISWC in SW mining session
libby: "an approach for automatic generation of frames out of tables which subsequently supports the automatic population of ontologies from table-like structures"
libby: html tables, that is
libby: paper presented at ISWC in SW mining session
libby: "an approach for automatic generation of frames out of tables which subsequently supports the automatic population of ontologies from table-like structures"
libby: html tables, that is
libby: by Phillip Lord, Sean Bechhofer, Mark D. Wilkinson, Gary Schiltz, Damian Gessler, Duncan Hull1, Carole Goble and Lincoln Stein4
libby: paper presented at ISWC in semantic web services session
libby: bioinformatics: distributed data, data-type intensive; currently you use html; service composition is copy and paste. Automation is via perl and screenscraping
libby: future solution - sws?
libby: the text formats produced are very complex, but strings only; no xml, big legacy problem
libby: 'thin semantics' - a simple wrapper to the text saying 'this is a blast file'"
libby: this can still help a lot with searching
libby: lots of data is represented in lots of says, e.g. 26 text formats for representing dna sequences
libby: use 'shims' pipe connectors - thin semantics again
libby: for services - can;t get any semantics from the wsdls, only from the websites; these user-orientated services have to be written by the users (e.g. pedro), could be a bottleneck - 600+ services
libby: ontologies are tricky without a lot of money, time and community involvement: do your own one (risk diverging form the biologist's descriptions); have an api where anyone can just add terms (seems to work quite well); use ontologies whereever you can find them (risks islands of interoprtability)
libby: I forgot to use quotes but these comments are paraphrased from the talk and slides
libby: paper presented at ISWC in semantic web services session
libby: bioinformatics: distributed data, data-type intensive; currently you use html; service composition is copy and paste. Automation is via perl and screenscraping
libby: future solution - sws?
libby: the text formats produced are very complex, but strings only; no xml, big legacy problem
libby: 'thin semantics' - a simple wrapper to the text saying 'this is a blast file'"
libby: this can still help a lot with searching
libby: lots of data is represented in lots of says, e.g. 26 text formats for representing dna sequences
libby: use 'shims' pipe connectors - thin semantics again
libby: for services - can;t get any semantics from the wsdls, only from the websites; these user-orientated services have to be written by the users (e.g. pedro), could be a bottleneck - 600+ services
libby: ontologies are tricky without a lot of money, time and community involvement: do your own one (risk diverging form the biologist's descriptions); have an api where anyone can just add terms (seems to work quite well); use ontologies whereever you can find them (risks islands of interoprtability)
libby: I forgot to use quotes but these comments are paraphrased from the talk and slides
Applying KAoS Services to Ensure Policy Compliance for Semantic Web Services Workflow Composition and Enactment
libby: by Andrzej Uszok, Jeff M. Bradshaw, Renia Jeffers, Austin Tate, Jeff Dalton
libby: paper presented at ISWC in semantic web services session
libby: more information here
libby: Coalition Search and Rescue - Task Support (CoSAR-TS), a related project
libby: "KAos provides necessary capabilities to verify and enforce user defined policy in the automated process of planning and executing workflows"
libby: "we use wordnet and googlesets to find synonyms for cell labels"
libby: paper presented at ISWC in semantic web services session
libby: more information here
libby: Coalition Search and Rescue - Task Support (CoSAR-TS), a related project
libby: "KAos provides necessary capabilities to verify and enforce user defined policy in the automated process of planning and executing workflows"
libby: "we use wordnet and googlesets to find synonyms for cell labels"
libby: paper presented at ISWC in semantic web services session
libby: a related document from the SWWS project
libby: "what is a service - 1) provision of value in some domain; 2) software entity to provide something of value; 3) a set of WSDL operations which implement some behaviours (webservice)"
libby: "3 phases: discovery, contract agreement, service delivery"
libby: a related document from the SWWS project
libby: "what is a service - 1) provision of value in some domain; 2) software entity to provide something of value; 3) a set of WSDL operations which implement some behaviours (webservice)"
libby: "3 phases: discovery, contract agreement, service delivery"
libby: 
jhendler:
-- found via image search on Google(tm) for monkey
jhendler: swoogle to the rescue - a monkey ontology (actually one of 13 classes found for "monkey" - yay swoogle!)
bengee: could be handy for the rdf world if it contained codemonkey (intersectionOf coder, monkey?)
jsled: Ooooh! A Monkey!
jhendler:
-- found via image search on Google(tm) for monkeyjhendler: swoogle to the rescue - a monkey ontology (actually one of 13 classes found for "monkey" - yay swoogle!)
bengee: could be handy for the rdf world if it contained codemonkey (intersectionOf coder, monkey?)
jsled: Ooooh! A Monkey!
libby: by Willem Robert van Hage, Maarten de Rijke, and Maarten Marx
libby: another paper from ISWC, in the rdf query section
libby: an approach to searching that uses both statistcs and ontologies
libby: another paper from ISWC, in the rdf query section
libby: an approach to searching that uses both statistcs and ontologies
logger: See discussion
libby: that url doesn't exist though....I guess later
crschmidt: That's the paper described in the papers.rdf
libby: that url doesn't exist though....I guess later
crschmidt: That's the paper described in the papers.rdf
Generating On the Fly Queries for the Semantic Web: The ICS-FORTH Graphical RQL Interface (GRQL) - abstract for rdfql section talk at ISWC
libby: by Nikos Athanasis, Vassilis Christophides, and Dimitris Kotzinos (who couldn't make it :(
libby: the full paper (pdf)
libby: isn't it great that lots of the papers are online despite not being on the site? :)
libby: looks like ithe paper will eventually be here (thanks crschmidt/julie!)
libby: demo!
libby: nice interface. uses the schema to generate the possible form-like queries by the looks of it
libby: the full paper (pdf)
libby: isn't it great that lots of the papers are online despite not being on the site? :)
libby: looks like ithe paper will eventually be here (thanks crschmidt/julie!)
libby: demo!
libby: nice interface. uses the schema to generate the possible form-like queries by the looks of it
libby: ISWC 2004 talk in rdf query section
libby: papers don't seem to be online (yet?)
libby: fortunately Peter Haase, the speaker/coauthor has put it all onlione anyway. other authors: Jeen Broekstra, Andreas Eberhart, Raphael Volz
libby: ericp asks a q about expressivity and logical entailment - separate or together? [dawg separates them, speaker says dependent; missed answer though...]
libby: question about why different serializations of RDF have not been considered in the survey: speaker says QL acts on the data model so not relevant
jeen: answer is that in Peter's opinion, formal semantics of the QL should be part of the specification of that QL and not be left to the implementation
libby: q from frank v-h - who are you so tolerant of qls that don't support rdfs? speaker: you can still be an rdf ql
libby: papers don't seem to be online (yet?)
libby: fortunately Peter Haase, the speaker/coauthor has put it all onlione anyway. other authors: Jeen Broekstra, Andreas Eberhart, Raphael Volz
libby: ericp asks a q about expressivity and logical entailment - separate or together? [dawg separates them, speaker says dependent; missed answer though...]
libby: question about why different serializations of RDF have not been considered in the survey: speaker says QL acts on the data model so not relevant
jeen: answer is that in Peter's opinion, formal semantics of the QL should be part of the specification of that QL and not be left to the implementation
libby: q from frank v-h - who are you so tolerant of qls that don't support rdfs? speaker: you can still be an rdf ql
DanC: found via ISWC tutorial materials by Pepper and Garshol
DanC: relative to earlier drafts by the same title, it seems to have many more contemporary references and some problems fixed, but the argument still seems flawed
danbri_: There is no crisis. Move along. Nothing to see here.
DanC: this version is fine until the section "Using the same URI to identify a different subject". To which the answer is, simply: don't do that.
danbri_: Talk of crisis is pure melodrama. RDF allows you to model things crudely, or with subtlety. The problem is not with RDF or the Web, but with particular schemas and datasets. So I resent the perpetuation of the FUD that we're in "crisis" because there are scruffy schemas and murky datasets in the Web.
logger: See discussion
DanC: "However, the TAG has no solution to offer, other than rather pathetic hand-waving". Hmm. speaking of dis-respect! I don't recall comments on webarch directly from Mr. Pepper.
Talliesin: Crisis, What Crisis?
Talliesin: We've been here before, and the best we can get from it is a third voice in the httpRange-14 debate that satisfies neither of the existing camps.
DanC: relative to earlier drafts by the same title, it seems to have many more contemporary references and some problems fixed, but the argument still seems flawed
danbri_: There is no crisis. Move along. Nothing to see here.
DanC: this version is fine until the section "Using the same URI to identify a different subject". To which the answer is, simply: don't do that.
danbri_: Talk of crisis is pure melodrama. RDF allows you to model things crudely, or with subtlety. The problem is not with RDF or the Web, but with particular schemas and datasets. So I resent the perpetuation of the FUD that we're in "crisis" because there are scruffy schemas and murky datasets in the Web.
logger: See discussion
DanC: "However, the TAG has no solution to offer, other than rather pathetic hand-waving". Hmm. speaking of dis-respect! I don't recall comments on webarch directly from Mr. Pepper.
Talliesin: Crisis, What Crisis?
Talliesin: We've been here before, and the best we can get from it is a third voice in the httpRange-14 debate that satisfies neither of the existing camps.
DanC: 7-11 November 2004 Hiroshima, Japan
DanC: thanks to libby et. al for ConnectingAudiences
DanC: animated GIFs. to laugh or to cry?
jhendler: or see the Semantic Metadata for the conference (Seamark(TM) viewer?)
DanC: some of the workshops and tutorials in the program look pretty interesting. Pointers to workshop notes appreciated.
DanC: any chance ISWC offers or endorses a wiki or blogroll?
libby: metadata for the papers - uses skos! (jeen says there's some (minor) character encoding problems though)
bengee: I wonder if the iswc2004 rdf schema is ever going to be re-used.. (ns uriref=http://iswc2004.semanticweb.org/ns#, no mappings to deployed terms, ..)
DanC: thanks to libby et. al for ConnectingAudiences
DanC: animated GIFs. to laugh or to cry?
jhendler: or see the Semantic Metadata for the conference (Seamark(TM) viewer?)
DanC: some of the workshops and tutorials in the program look pretty interesting. Pointers to workshop notes appreciated.
DanC: any chance ISWC offers or endorses a wiki or blogroll?
libby: metadata for the papers - uses skos! (jeen says there's some (minor) character encoding problems though)
bengee: I wonder if the iswc2004 rdf schema is ever going to be re-used.. (ns uriref=http://iswc2004.semanticweb.org/ns#, no mappings to deployed terms, ..)
First keynote ISWC speaker: Edward Feigenbaum (the text below is paraphrased from his talk and slides. apologies for innacurracies etc)
libby: Feigenbaum collaborated with Jim Hendler a while back on a paper, maybe this one
libby: searching around he found slideshows but not killer apps
libby: he interviewed sw gurus. is the sw 10 x better than google? is the benefit/cost ratio large? logic is a burden (complex, slow)
libby: revisiting 70-80s AI work - though AI and SW is not isomorphic, there's a big overlap
libby: the SW as web scale expert system(s)
libby: we live with systems which are not 100% correct: google!
libby: "does logic give us the best set of tools for the practical sw? the evidence is mostly lacking"
logger: See discussion
libby: "build a little, test a little" "find a task domain that has a real world importance, is rich with posibilities, then do a series of experiments, ascending the sprial"
jhendler: the paper libby mentions is at this springer link
libby: this community has a tendency to work top-down rather than bottom up"
jhendler: which takes you to a download page.
libby: he talks about example applications - to help you answer archival questions: Feigenbaum suggets stuff like - how did X meet Y to collaborate on the Z project (in science), and other interesting questions
libby: "cybergraves" for scientists..and also the non-famous
DanC: among the ISWC invited speakers
libby: "when a distinguished and elderly scientist says somethign is possible he is almost certainly correct; when he says something is impossible, he is almost certainly wrong" - Arthur C clarke
libby: "stop so much talking and start more doing"
libby: searching around he found slideshows but not killer apps
libby: he interviewed sw gurus. is the sw 10 x better than google? is the benefit/cost ratio large? logic is a burden (complex, slow)
libby: revisiting 70-80s AI work - though AI and SW is not isomorphic, there's a big overlap
libby: the SW as web scale expert system(s)
libby: we live with systems which are not 100% correct: google!
libby: "does logic give us the best set of tools for the practical sw? the evidence is mostly lacking"
logger: See discussion
libby: "build a little, test a little" "find a task domain that has a real world importance, is rich with posibilities, then do a series of experiments, ascending the sprial"
jhendler: the paper libby mentions is at this springer link
libby: this community has a tendency to work top-down rather than bottom up"
jhendler: which takes you to a download page.
libby: he talks about example applications - to help you answer archival questions: Feigenbaum suggets stuff like - how did X meet Y to collaborate on the Z project (in science), and other interesting questions
libby: "cybergraves" for scientists..and also the non-famous
DanC: among the ISWC invited speakers
libby: "when a distinguished and elderly scientist says somethign is possible he is almost certainly correct; when he says something is impossible, he is almost certainly wrong" - Arthur C clarke
libby: "stop so much talking and start more doing"
JibberJim: The example HTML links are broken!
JibberJim: references http://www.w3.org/RDF/icons/rdf_developer_button.64
balbinus6: W3C's link checker says 301 for Eric Miller's homepage
balbinus6: Thanks mortenf for the SVG (by-sa 2.0, btw)