oor_chat-transcript_20100219e_edited.txt PeterYim: . Welcome to the OOR Panel Discussion Session - Fri 19-Feb-2010 Title: Coordinating our Open Ontology Repository Software Development Session Chair: MikeDean (Raytheon BBN) Panelists: * JimChatigny (Raytheon BBN) * MichaelGruninger (University of Toronto) * KenBaclawski (Northeastern University) * NatashaNoy (NCBO / Stanford University) Please refer to details on the session page at: http://ontolog.cim3.net/cgi-bin/wiki.pl?OOR/ConferenceCall_2010_02_19 . anonymous1 morphed into MishaDorf anonymous1 morphed into JimDisbrow PeterYim: another plea (while we are waiting) ... please hep with our OntologySummit2010 Surveys - see: http://ontolog.cim3.net/cgi-bin/wiki.pl?OntologySummit2010_Survey PeterYim: take a look at the solicitation too, at: http://ontolog.cim3.net/cgi-bin/wiki.pl?OntologySummit2010_Survey_Solicit anonymous morphed into PaulAlexander anonymous1 morphed into LinePouchard anonymous morphed into JackPark LinePouchard: we're not funded yet, but am working on http://dataone.org JimDisbrow: @Line: consider joining us - go to http://ET.gov, on the left is a "Search" button, click it and the next screen will show a list of ET.gov components - one of which is "Energy-Water Nexus vis-a-vis the Climate"; click on the referenced XML file and it will lead you into what the project is doing anonymous morphed into BartGajderowicz anonymous1 morphed into DougFoxvog anonymous2 morphed into LinZhang (Forest) DavidEddy: I see you here JimDisbrow RaviSharma: Hi Jim JimDisbrow: hi Ravi, David DougFoxvog: Could anonymous3 click "settings" and put in their name? anonymous morphed into MattHettinger DavidEddy: SCM (software conf mgmt) is EXTREMELY important around repository... a fatal missing component over past 40 years LinZhang (Forest): To Session Chair: I'm LinZhang (Forest) from China. And I'm on. PeterYim: @JimChatigny - I noticed on your slide 4, you were federating from the OOR-sandbox at ... please note that the OOR-sandbox is now at http://oor-01.cim3.net (the name was for testing only ... for everyone's information) RaviSharma: @MikeDean and JimChatigny: Is there an architecture diagram to look at that shows pattern or Ontology Beans, etc? or did I miss it? PaulAlexander: You can view a complete description of NCBO REST services (which at this point match OOR) here: http://www.bioontology.org/wiki/index.php/NCBO_REST_services PeterYim: @Ravi - guess you are aware of this: http://ontolog.cim3.net/cgi-bin/wiki.pl?OOR/ConferenceCall_2010_02_19#nid2955 ... maybe Jim could work with Todd and consider adding to it RaviSharma: Peter: great link and lot of architecture information in UML that ToddSchneider developed that I had missed so far. Thanks anonymous2 morphed into FrankOlken TerryLongstreth: Have you trained anyone to administer the system in your absence? JimRhyne: Is there an introduction and rationale for this particular OOR architecture? Not obvious to me why Spring would be a good basis. PaulAlexander: @JimRhyne: Misha may be able to speak to the decision to use Spring. It's in OOR because it is used in NCBO BioPortal (which OOR is an instance of right now) JimRhyne: @MishaDorf - can you post a link to information on NCBO BioPortal architecture and implementation? PeterYim: @JimRhyne - see resources listed under: http://ontolog.cim3.net/cgi-bin/wiki.pl?OOR/ConferenceCall_2010_02_19#nid2950 PaulAlexander: Here's a BioPortal/OOR architecture overview: http://www.bioontology.org/wiki/index.php/NCBO-OOR_Architecture JimRhyne: @PaulAlexander - exactly what I was looking for, thanks. anonymous morphed into DanCerys Natasha Noy: @JimChatigny Is there a link where we can try the federated server? Natasha Noy: Also @JimChatigny: I think I missed this: do you download only metadata, or the ontologies themselves as well? Natasha Noy: And Q3 @JimChatigny: Is there a way to limit what to federate from a particular repository (e.g., only OWL ontologies)? RaviSharma: @JimChatigny: Are you restricted to RDF? especially due to REST requiring a resource? RaviSharma: @MichaelGruninger: What about intermodule integration? Does CL fully prescribe a unique way of putting them together? RaviSharma: @MichaelGruninger and AliHashemi: Hope you are aware of current progress being made by Mike Linehan and others in our efforts at OMG on SBVR Data-Time? RaviSharma: @MichaelGruninger and AliHashemi: These have time and duration vocabularies that would be helpful in developing ontologies of temporal aspects? LinZhang (Forest): @MichaelGruninger: Did the current use case include the natural language translation of ontologies? If yes, maybe we can use the term "converter" rather than "translator" for the syntax conversion. What do you think of this? Thanks. PatCassidy: @MikeGruninger: Do you think that JohnSowa's idea of using OpenCyc as the base for relating multiple ontologies could be implemented in COLORE? Any ideas on what such an exercise would need in terms of work.resources? MichaelGruninger: @PatCassidy: We have found that we do not need to use any ontology as an a priori "base for relating multiple ontologies". Once the axioms of the ontologies have been explicitly stated, their relationships are much easier to identify. The problem is that nobody bothers to specify their axioms when discussing the relationships between ontologies. One motivation for COLORE was to get people to actually submit axioms. PatCassidy: @MikeGruninger: Yes, but axioms can only be related to each other if the namespaces are identical, or if some translation is provided. The issue I am concerned with in relating ontologies is to specify the bridging axioms that convert the assertions in one domain ontology into the equivalent assertion in the other ontologies. So if the 25,000 - 60,000 OpenCYc hierarchy is loaded and axioms are added, how do you imagine this will interact with the other modules in COLORE? DougFoxvog: @MikeGruninger: without a base for relating ontologies, axioms expressed in one ontology could not be interpreted in another. Relationships between them could not be determined without mappings expressed. AliHashemi: @DougFoxvog - can you elaborate what you mean by "base" ? DougFoxvog: I used MikeGruninger's term.. MichaelGruninger: @DougFoxvog: If one ontology is interpretable by another, we don't need a "base ontology" to show this. You simply specify a mapping from the lexicon of one to the lexicon of the other and then show that theorems are preserved. DougFoxvog: there should be some way to specify a mapping betweens terms. A star mapping is much simpler than creating an N-N mapping; but the center MichaelGruninger: BTW, the term "base ontology" is PatCassidy's and John Sowa's. DougFoxvog: of the radial star might be more than one ontology AliHashemi: @DougFoxvog - one need not create N-N mappings. If you map A to B and perhaps map B to C, then you have an implicit mapping, from A to C through B. The star structure emerges. DavidEddy: I'd be interesting in talking with anyone who's interested in examining the lessons learned from 40 years of repository efforts... 95% of which have gone to the bit bucket in the sky. WHY did they fail & what can we avoid repeating? PeterYim: @KenBaclawski- (slide#6) what is KEEPER? is it s software tool? if so, is it an opensource software? JimRhyne: @PeterYim phone battery died, back on now. PeterYim: @JimRhyne - glad you're back RaviSharma: @KenBaclawski: I believe you are simply executing the OOR gatekeeping function without touching the content or relationships or internal structure of ontology, ie. using BPEL for that purpose? KenBaclawski: @Peter: KEEPER was the name of the project in my course. It is a web service (supporting WSDL, SOAP and REST) and will be open source. TerryLongstreth: @KenBaclawski - you mention 11179. Are you familiar with XMDR? Has it provided anything useful? KenBaclawski: @Terry: Yes, we did look at XMDR and we hope to use some of it in KEEPER. LinZhang (Forest): @TerryLongstreth: XMDR is a good resource. In some way, OOR is like XMDR. PeterYim: we actually have XMDR folks on the OOR team (notably BruceBargmeyer ... although he has been less active lately) KenBaclawski: @Peter: I have been in contact with the XMDR people, but not much has occurred in this direction. LinZhang (Forest): @MichaelGruninger: Do the current use cases include the natural language translation of ontolgies? Thanks. AliHashemi: @Lin, as far as I know, the current use cases do not include NL translation of ontologies. LinZhang (Forest): Years ago, I tranlated some wiki pages into Chinese and post there. LinZhang (Forest): @AliHashemi: Thanks. But the feature is desirabe. KenBaclawski: @Ali, the use cases are still being developed. If you are interested in this, please come to the Use Cases meeting. LinZhang (Forest): I uploaded some ontologies with Chinese concept names onto the BioPortal, but it couldn't display the Chinese character correctly. PaulAlexander: @LinZhang: We recently discovered a bug that was preventing these characters from displaying properly. BioPortal now displays them properly, but the OOR Sandbox may be running code from prior to the bug fix. LinZhang (Forest): @PaulAlexander: Oh, that's great! I'll check that later after the t-con. Thanks. LinZhang (Forest): Amino Acid with Simplified Chinese annotations (View for Amino Acid) Version 1.3: http://bioportal.bioontology.org/visualize/41006?view=true LinZhang (Forest): @PaulAlexander: Thanks. It works. Amino Acid with Simplified Chinese annotations (View for Amino Acid) Version 1.3: http://bioportal.bioontology.org/visualize/41006?view=true Biomedical Resource Ontology with Simplified Chinese annotations: http://bioportal.bioontology.org/ontologies/40656#views AliHashemi: @DougFoxvog, slide 4 in MichaelGruninger's presentation provides a high level view of "some way to specify a mapping between terms" DougFoxvog: The star model is what JohnSowa was suggesting for the FO (Foundation Ontology). That was also one of original ideas behind Cyc. A center of the star might have multiple inter-related ontologies (SUMO, Cyc, etc.). Sowa is considering an FO w/ multiple modules (for different topics). AliHashemi: @DougFoxvog, the problem with top-down star models are exactly that, they are top down. Another approach is to have them emerge from the mappings that exist between theories. For example, mathematical concepts (i.e. geometries, orderings) provide the backbone for many many ontologies. They serve in many was a central star. Of course, one doesn't know unless axioms are inputted, and mappings are generated. DougFoxvog: On slide 4, pi specifies a mapping function from T0 to T1. This does not explain whether Pi0-1 is calculated via a star, N-N, chain of mappings, or any other specific mapping. PatCassidy: @AliHashemi: If you map all of A to all of B, there are likely to still be terms in one that are not in the other and vice versa, so there will be terms not mapped. Then if you map C to B, some of the terms in A (that have not been mapped to B because they don't exist in B) will not be mapped from B to C for the same reason. The transitivity will only be complete by this mechanism if all ontologies have the same set of terms. One can use this case-by-case mechanism, but if you keep a "foundation" ontology that contains all of the terms that have been mapped in common between any two ontologies, then the foundation ontology will be able to serve as a reliable and complete intermediate for translation among all of the ontologies in the repository. That is the principle I have suggested for the FO; it is not disjoint from your approach, it just *adds* an FO to serve as the most complete inventory of terms needed for translation. MichaelGruninger: @DougFoxvog: The definition of interpretation (on my slide 4) does not assume any algorithm for generating the mapping. Ali's thesis and the current paper discuss ways of generating new mappings from the relationships between ontologies that are specified in the repository. AliHashemi: @PatCassidy, a partial mapping will be a partial mapping, regardless of whether it is into a foundation ontology or not. You can create a new module which enables a full mapping (i.e. extend the FO), but then we are moving away from an FO. Note you could also do this for A to B in the above example. ... The goal you have outlined is laudable, I've already voiced some concerns. You are right however that attempting to identify ontological primitives is complementary to much work being done and could help in the generation of mappings. BartGajderowicz: @Pat, @Ali: Pat as you mentioned getting every single term in A to match every term in B is very unlikely, simply because A and B were created separately. Could Ontology Granulation assist in this, where it identifies key low-level terms (properties) which may be sufficient to *differentiate* higher level terms (concepts). Then these partial mappings will have been chosen based on *best* terms from the limited number of available terms. *best* terms may introduce a level of stability in the partial mappings. AliHashemi: Bart, that might be a useful approach in identifying what Pat calls primitives. I would note that before one can identify primitives, one needs to _at least_ specify mappings between some of the dominant upper ontologies, and i would suggest, mathematical theories. PatCassidy: @BartGajderowicz: The Foundation Ontology as I envision it would be an ontology that is mapped (by complete translations) to every other ontology in the repository. So if terms in different ontologies have some conceptual overlap, but are not identical, the part that overlaps will be identifiable via the FO. This FO could be created by performing the kind of mappings that Ali suggests, but also performing a mapping to the FO. When there are terms needed that are not in the FO, they can be added. But the additions to the FO are kept to a minimum by only adding the most basic terms (primitives) that are required to logically specify the terms in all of the mapped ontologies. This mechanism would guarantee an automatic translation of assertions in any ontology into assertions in any other mapped ontology. DougFoxvog: @MikeGruninger: Re your page 10, without a base for relating ontologies, axioms expressed in one ontology could not be interpreted in another. Relationships between them could not be determined without mappings expressed. For computers, a ROT-13 of the names of one ontology would make no difference. DougFoxvog: OOPS. Ignore my last comment. I meant: DougFoxvog: @MikeGruninger: Re your page 10, By having ontology terms for modules as instances, common logic could be used to express simple relationships among ontologies. The issue would be whether Common Logic can associate such terms with the modules themselves. DougFoxvog: @PatCassidy, do you envision a functional description of terms in the FO for mapping to more specific terms. E.g., (NthInSeriesFn (AnatomicalPartTypeFn (LeftFn Hand-AnatomicalPart) FingerSeries) 3) for a SNOMED term for the third finger of the left hand? PatCassidy: @DougFoxvog: The FO only needs to have the basic elements that can be combined by FOL to produce the terms in the domain ontologies. So Nth in series might be a primivie, but applied to the hand it could be a domain-specific comnbination. DavidEddy: can someone point to a ontology glossary... I have no idea what a "theory" or "axiom" is please? LinZhang (Forest): @PeterYim and MikeDean: I'll be out about 20 min later because my skype card is running out. PaulAlexander: Is anyone familiar with/using Git SCM? It could work very well for this type of distributed development. PaulAlexander: Or Mercurial or some other distributed SCM system RaviSharma: @KenBaclawski: My Question is for you from above relating to ISO 11179 that defines metadata Registry and data element definitions hence the question, does OOR automatically assume that things and relationships are "data Elements"? Also BPEL Q from Above repeated here for you.I believe you are simply executing the OOR gatekeeping function without touching the content or relationships or internal structure of ontology, ie. using BPEL for that purpose? KenBaclawski: @RaviSharma: ISO 11179 can specify content and relationships. However, the granularity of administered item for OOR remains to be determined. This is related to modularity. The administered items are the units being managed with a BPEL process. LinePouchard: @KenBaclawski: Is your NIH project planning to develop a new "environement" ontology? Do you have a list of existing ones? KenBaclawski: @LinePouchard: We will be using existing ontologies when possible. Send me email. LinePouchard: Ken, Peter and all: I have to go. Ken: I will be in touch by email. Thank you. PeterYim: on code repository - MikeDean, on consultation with NatashaNoy and everyone at this call: decision reached (confirmed between the OOR-team and NCBO-team) - OOR will be using the repository at semwebcentral for the OOR-branch (of the BioPortal extensions) located at: http://oor.projects.semwebcentral.org/ - ref: http://ontolog.cim3.net/cgi-bin/wiki.pl?OpenOntologyRepository#nid1JSY JimRhyne: Unfortunately have to leave the call. DougFoxvog: Signing off at 6:30 p.m. (UTC), i need to head home from work. PeterYim: thanks for joining us Doug RaviSharma: Thanks Mike Peter and Speakers. AliHashemi: Thanks all. Bye PaulAlexander: Thanks everyone, please let me know if you have questions as you move the OOR codebase to the new SVN repository. PeterYim: - session ended 2010.02.19 - 10:31am PST -