Chat History: Nanopublications

Created on 2011-02-16 17:05:11.

David R Newman: 14:40:27; Hi Sean. I assume this nanopublication/swan/research objects chat is at 16:00 GMT?
Sean Bechhofer: 15:05:56; yes -- I'll call you.
David R Newman: 15:06:29; Ok, just so I can be somewhere quiet for 4pm
Sean Bechhofer: 15:54:43; Will call you all in a couple of minutes.
Paolo Ciccarese: 15:59:01; ok
Sean Bechhofer: 16:03:02; are you there david?
Paolo Ciccarese: 16:09:57; i lost you
Sean Bechhofer: 16:10:07; restarting
Sean Bechhofer: 16:13:40; What are the success criteria?
Sean Bechhofer: 16:14:15; Paul: SHould be able to take two nano-publications that talk about the same thing and remove redundancy (but maintain provenance).
Sean Bechhofer: 16:14:46; Paul: If it's the same triple but with different provenance.
Sean Bechhofer: 16:15:23; Paul: Once we remove redundancy can say there are 10,000 instances of this nano-pib
Sean Bechhofer: 16:15:35; Paolo: Lots of inconsistency in the database.
Sean Bechhofer: 16:16:01; Paolo: Identfying conflicting statements
Sean Bechhofer: 16:16:27; Sean: What's conflict?
David R Newman: 16:16:42; Hi Sean
David R Newman: 16:16:48; ready now
Sean Bechhofer: 16:17:53; Sean: How to identify conflicting triples
Sean Bechhofer: 16:18:09; paul:Avoid context, but encode statements with provenance.
Sean Bechhofer: 16:18:57; Paul: Can then detect statements with more, less or joint provenance. Don't model conflict or context. Expose things with lots of provenance.
Sean Bechhofer: 16:19:24; Sean: Don't try and answer the question. Punt on it. "Here's a bunch of stuff, do what you want with it".
Sean Bechhofer: 16:20:01; Paolo: Taking facts out of context don't mean anything.
Sean Bechhofer: 16:20:21; Paul: Where do embed the context. It's given by its provenance.
Sean Bechhofer: 16:20:56; Paul: Provenance includes a source document.
Sean Bechhofer: 16:21:00; Paul: OR a source db
Sean Bechhofer: 16:21:22; Paolo: Don't encode the context, point at it and allow the consumer to interpret it.
Sean Bechhofer: 16:22:00; Paul: Would rather have someone else worry about context. Anita de Waard tries to do that. With nanopubs we avoid that.
Sean Bechhofer: 16:22:31; Paolo: You never have in Alzheimers a statement that's a triple.
Sean Bechhofer: 16:22:51; Paolo: THings like "a large amount of <some stuff> is toxic for the brain".
Sean Bechhofer: 16:22:57; So it's not at a triple level.
Sean Bechhofer: 16:23:08; Paul: Can encode that as a mini[graph.
Sean Bechhofer: 16:23:23; That's reasonable. Want to avoid mini grpahs that become large.
Sean Bechhofer: 16:24:05; Sean: When is too big too big?
Sean Bechhofer: 16:24:54; Paolo: Rule of thmb. Curators create statements. As simple as they can. But more than triples.
Sean Bechhofer: 16:25:11; Paul: Can encode more into the URLs.
Sean Bechhofer: 16:26:44; Sean: What do you need for claims other than an identifier?
Sean Bechhofer: 16:31:51; Sean: So claims need to have 1/ a means of identification and determining equivalence; and 2/ some means of search/discovery that may require ooking inside the claim.
Sean Bechhofer: 16:32:26; and 3/ not too big/complex
Sean Bechhofer: 16:35:07; Paul: Can't you split the claim into a triple?
Sean Bechhofer: 16:35:18; Paolo: Need to make a load of stuff up.
Sean Bechhofer: 16:35:39; Paolo: Can agree on provenance, but not on content.
Sean Bechhofer: 16:36:02; Paolo: Could aske Nigem Shah to encode this using ontologies already existing.
Sean Bechhofer: 16:37:30; Sean: Worried about forcing claims into triples.
Sean Bechhofer: 16:37:45; Paolo: hard to define a self-contained claim and then hard to formalise.
Sean Bechhofer: 16:38:30; David: What's important is finding concepts so that things can be reused. Also when a new one is created, need to provide sufficient description so that someone else can find it.
Sean Bechhofer: 16:38:46; David: End up with multiple creation of similar things.
Sean Bechhofer: 16:39:21; Sean: Dont't you end up with lots of stuff not interlinked?
Sean Bechhofer: 16:39:40; Paul: People making claims and claim from data that's there.
Sean Bechhofer: 16:39:56; Paul: Would be interesting to look at nano-pubs from data sets. Would be cleaner.
Sean Bechhofer: 16:40:08; Paolo: Not so sure. Quantittive.
Sean Bechhofer: 16:40:18; Paul: Can use RDF for that but it's a pain.
Sean Bechhofer: 16:41:02; Paul: Will try and find a nano-pub based on data. Use the claim in Paolo's example.
Sean Bechhofer: 16:41:23; Sean: Doesn't that involve interpretation?
Sean Bechhofer: 16:42:11; Paulo: Let me try and select something simpler that can more easily be found in data sets.
Paul Groth: 16:43:44; tumor protein p53 has phenotype Colon_cancer
Paul Groth: 16:44:12; tumor protein p53 has bio marker UniSTS:158212
Paul Groth: 16:44:28; http://linkedlifedata.com/resource/entrezgene/id/7157
Sean Bechhofer: 16:45:29; Paul: statements made here don't have provenance.
Sean Bechhofer: 16:46:03; Paolo: Notion of canonical claim. Lots of claims that were slightly similar but phrased differently. Biologists would say "they're the same thing".
Sean Bechhofer: 16:46:13; Paolo: Idea of canonical statement.
Sean Bechhofer: 16:46:50; Paolo: Statement tumor protein.... is the tip of the iceberg.
Sean Bechhofer: 16:47:34; Paul: Like idea of the canonical statement.
Sean Bechhofer: 16:48:12; Paul: Just saying things without provenance is less interesting.
Sean Bechhofer: 16:48:49; ACTION: Paolo talk to Nigem (sp?)
Sean Bechhofer: 16:48:58; ACTION: Paul to look at data nano-pubs
Sean Bechhofer: 16:49:12; ACTION: Get this coded up in a triplestore. [Paul?]
Paolo Ciccarese: 16:49:28; Nigam H. Shah
Sean Bechhofer: 16:50:33; Paolo: Can produce RDF triples but not named graphs.
Paul Groth: 16:50:58; http://www4.wiwiss.fu-berlin.de/bizer/TriG/
Sean Bechhofer: 16:52:24; Sean: What do we ask once we've got it in a t-store?
Sean Bechhofer: 16:52:38; Paul: Find me all the informaiton about myelation (and provenance)
Paolo Ciccarese: 16:52:56; HyBrow A prototype system for computer-aided hypothesis evaluation - http://www.hybrow.org/
Sean Bechhofer: 16:53:00; Paul: Removing redudancy, but none in there right now. Could fake some.
Sean Bechhofer: 16:53:40; Paolo: Could ask someone from Alzheimers Knowledge Management.
Sean Bechhofer: 16:53:51; Paolo: Would give deeper queries.
Sean Bechhofer: 16:54:26; Paul: Find me all things confirmed in two different publications. Query about facts related to their provenance
Paolo Ciccarese: 17:01:18; http://www.w3.org/wiki/HCLSIG/SWANSIOC
Sean Bechhofer: 17:02:03; ACTION: Paolo to create a page on the HCLS wiki.
Sean Bechhofer: 17:03:33; Paul: Can these be used as stand off annotations for ROs to improve search.
Paolo Ciccarese: 17:03:53; I am out again :/
Paul Groth: 17:04:03; we're just wrapping up
Paolo Ciccarese: 17:04:10; what is RO?
Sean Bechhofer: 17:04:21; Research Object
Paolo Ciccarese: 17:04:25; oh ok
Sean Bechhofer: 17:04:33; Bye all!
Paolo Ciccarese: 17:04:36; bye :)
Paolo Ciccarese: 17:04:51; Sean, are you going to send all of us the notes?