ppy/oor_chat-transcript_unedited_20120814a.txt --------- Chat transcript from room: oor_20120814 2012-08-14 GMT-08:00 --------- [08:22] PeterYim: Welcome to the = "OOR for Big Data" Workshop-I - Tue 2012_08_14 = Topic: "OOR for Big Data" Brainstorm Session Session Chair: Mike Dean (OOR; Raytheon-BBN) Session page: http://ontolog.cim3.net/cgi-bin/wiki.pl?OOR/ConferenceCall_2012_08_14 Mute control: *7 to un-mute ... *6 to mute Can't find Skype Dial pad? * for Windows Skype users: Can't find Skype Dial pad? ... it's under the "Call" dropdown menu as "Show Dial pad" * for Linux Skype users: please stay with (or downgrade to) Skype version 2.x for now (as a Dial pad seems to be missing on Linux-based Skype v4.x for skype-calls.) == Proceedings: == [08:37] PeterYim: == MikeDean starting the session off with his intro slides ... [08:41] ToddSchneider: How doe these supporting technologies fit into overall IT/data architectures? [08:47] PeterYim: == Q&A and Open Discussion ... [08:48] anonymous morphed into ElizabethFlorescu [08:49] TerryLongstreth: Mike mentioned RDF from SQL productions; Terry asked if that included capturing the logic from constraints, triggers (Mike added views). [08:51] MikeBennett: How do you manage the relationship between the logical data model which corresponds to the (e.g. SQL) data source, and the ontology of the domain. How do we distinguish an "ontology" which is an RDF model of a data model design, versus an ontology of the real things? [08:53] MikeBennett: That is, how to formalize the model theoretic relationship between the elements in the RDF model, and the things which they represent (data elements v things). [08:54] PeterYim: our opportunity (and challenge, of course) could be in taking where the Linked Data people leave off (say, having extracted a vocabulary on the dataset) and come up with the ontology and the value added services that OOR is poised to provide ... of course, we should look at starting from the dataset as well (not just from the vocabulary) [08:58] MikeBennett: Suggestion: in extracting information from large datasets, is there a role for an "Ontology of the data" and a clear way of distinguishing this from the ontology of the subject matter itself? [09:01] MichaelGruninger: @MikeBennett: Do you see a role for an "ontology of the data"? What kinds of concepts and relations do you see in such an ontology? Would it be domain-independent? [09:03] PeterYim: candidate data that we might get our hands on and communities we can collaborate with are probably in the domains of: (i) government data; (ii) geospatial / geo-science; (iii) biomedical [09:04] MikeBennett: This is what I'm wondering. It's all too easy for someone to take a logical model (which has been designed, rightly), converting it into RDF/OWl and saying "Lo, an ontology". I think there must be a role for it in data extraction, but also a clear difference to ontologies of the subject matter. With perhaps an RDF/OWL mapping ontology between the two? [09:05] PeterYim: a mini-series is coming up shortly - see: http://ontolog.cim3.net/cgi-bin/wiki.pl?EarthScienceOntolog ... the kick-off session for the series is coming up next week (Thu 2012.08.23) [09:06] MikeBennett: @Michael PS I suspect that the basic concepts such a model would be built from, i.e. the top level classes, might be the constructs that exist in the data modeling language, e.g. data table, join table, UML class, UML attribute and so on? I don't know, I'm still ruminating on this. [09:06] TerryLongstreth: Big data characterized by Size, by Complexity, by SxC? [09:08] TerryLongstreth: A characteristic of big systems is the fuzziness of the boundaries; does the same hold for big data? [09:10] ToddSchneider: Ken, when the phrase 'big data' is used I haven't seen qualification to the number of sources. Perhaps there's an implicit assumption of a single source. [09:11] KenBaclawski: When we talk about Big Data and OOR, there are two different problems: a relatively small number of large complex ontologies (as in BioPortal which has ontologies with millions of concepts) or a very large number (on the order of millions) of relatively simple ontologies (e.g., the fields of a CSV file). [09:12] PeterYim: == focused discussion on: Supporting Technologies [09:18] TerryLongstreth: Challenge for OOR would seem to be more towards complexity rather than size of a Big Data environment [09:19] ToddSchneider: Ken, Okay there may be performance issues when the number of ontologies becomes large. [09:20] MikeBennett: Open linked data marketing issue: so all these governments that are putting out RDF linked data on the basis of the benefits of semantics: are they aware of the available ontologies (per OOR) so that they start to use existing ontologies as the conceptual model from which to structure new and future open linked data outputs. [09:20] MikeBennett: This also requires an awareness of the role of an ontology AS a conceptual model - many folks are simply not aware of basic top down modeling best practice. [09:22] MichaelGruninger: Two different kinds of applications of OOR: 1) using ontologies within OOR together with existing data sets e.g. integration, decision support. 2) using ontologies within OOR to design new data sets or redesign existing data sets [09:27] MikeDean: As part of "moving to the cloud", many organizations are moving from relational databases to map/reduce frameworks such as Hadoop. Can ontologies be used to assist with or guide such migration? [09:30] ToddSchneider: Have to go. Cheers. [09:34] PeterYim: ^candidate data: ... (iv) standards [09:37] KenBaclawski: Big Data also involves format conversion issues as well as semantics (e.g., integer vs floating point, image formats). [09:39] MikeBennett: OOR: would not include metadata about data formats (how could it?) but there is a role for some demonstration of description of how to use this stuff, as part of the usage of OOR in that scenario. [09:39] KenBaclawski: Clarify these terms: data registry, data repository, ontology registry, ontology repository. [09:40] MikeBennett: Also clarify how these different moving parts may be framed in terms of the Zachmann Framework or some similar formal development framework. [09:40] TerryLongstreth: The loss of meaning in data conversions should be part of the ontology [09:41] TerryLongstreth: ...perhaps a separate micro-theory about data formats and lossy vs. lossless conversions [09:44] MikeBennett: Incompatible data sources: what ontologies are for. [09:47] MikeDean: http://icpsr.umich.edu is a large repository of social science data [09:47] KenBaclawski: I suggest as an action item: capture the use cases we just identified. [09:48] MikeDean: Most domains/projects seem to have their own data registries [09:58] TerryLongstreth: Are agreed that Big Data for OOR is primarily an issue of complexity of the target relationships? [09:59] TerryLongstreth: i /are we/ [10:00] TerryLongstreth: Apparently no consensus yet [10:01] MikeBennett: Or is it "Data that is Big" v "Big data architecture" i.e. Hadoop / Mapreduce. Both seemed relevant according to today's conversation. [10:04] PeterYim: as long as we are not arguing about what "Big Data" is ... and we know our objective is to tackle how OOR can "serve" Big Data, I think we are fine (at least for now; until we need to tackle at finer granularity) [10:04] PeterYim: great session! [10:05] PeterYim: -- session ended: 10:03am PDT -- [10:12] PeterYim: ^assuming we have consensus on putting a focus on "OOR for Big Data" (along with what we have been doing so far with OOR), we should: (a) identify the data, use case(s) and one or two community(ies) we want to work closely with to take this effort to the next level, (b) figure out what we need to be doing differently (from this point onwards) for OOR [10:21] PeterYim: Following event dates confirmed [10:21] PeterYim: == Next Meetings: ** Tying down event dates - ref. http://ontolog.cim3.net/cgi-bin/wiki.pl?OOR/ConferenceCall_2012_08_07#nid3DJ5 *** OOR Architecture & API Workshop-XIII - Co-chairs: KenBaclawski & ToddSchneider - "use cases" - Tue 2012_09_11 *** OOR Metadata Workshop-VIII - Chair: MichaelGruninger - will do "Mapping" and/or "metadata in OOR for Big Data" - Tue 2012_09_25 *** OOR Funding-I - Chair: KenBaclawski - "rethink strategy" - Tue 2012_08_21 (MikeBennett voiced his preference for 8/21; most people are ok with either dates.) *** OOR Code Development-IX - Chair: MikeDean - Tue 2012_09_16 *** OOR Infrastructure-II - Chair: PeterYim - tba ** No meeting on Tue 2012_08_28 ---------