Joint OpenOntologyRepository-OntologySummit2008 Panel Discussion Session - Thu 27-March-2008    (1C1T)

Conference Call Details    (1C28)

Attendees    (1C33)

Background    (1C3H)

Two parallel initiatives are ongoing in the community, revolving around the theme of "Open Ontology Repository". On the one hand, a working group under the auspices of the OpenOntologyRepository Initiative, and on the other, the discourse (and essentially a discussion group that culminates in a two-day workshop) conducted as the main focus for OntologySummit2008.    (1C3I)

It is at the intersection of these two initiatives that this panel discussion session is being held. The OpenOntologyRepository team is taking the opportunity to have some of its members who are bringing technology and infrastructure to the table to present them side-by-side, and to discuss how these can all fit nicely together. The OntologySummit2008 folks, on the other hand would want to take the opportunity to survey (at least a subset of) the technology & infrastructure landscape to gain insight into the state-of-art in Ontology Registry and Repository.    (1C3J)

Besides hearing from the panelists, we are setting aside ample time after their briefings, for some good Q&A and discussions among all who are participating in this session.    (1C3K)

Refer to details at the respective project homepages of the two initiatives at: http://ontolog.cim3.net/cgi-bin/wiki.pl?OpenOntologyRepository . & . http://ontolog.cim3.net/cgi-bin/wiki.pl?OntologySummit2008    (1C3L)

Agenda & Proceedings:    (1C3M)

Title: An Open Ontology Repository: Rationale, Expectations & Requirements    (1C3U)

Abstracts:    (1C3V)

Resources    (1C4H)

Questions, Answers & Discourse:    (1C55)

Questions and Discussion captured from the chat session:    (1C5D)

 RaviSharma: Q For WilliamBug: How is the usage of BIRN work by 1. research 2. operaional practitioners 
            progressing especially in adoption of common or cross discipline concepts and vocabularies 
            and Other collaborative features?    (1C5E)
 LeoObrst: Question to BillBug: Biomedicine is changing so quickly. How do you integrate working hypotheses 
          (not yet firm) with fairly established theories (ontologies)? Is this very fast changeability a 
          hard problem? How do you keep up?    (1CIS)
 RaviSharma: Q for EvanWallace, can we think of a combined collaborative OOR Repository &/or Registry 
            definition or Repository &/ or Registry implementation RFPs in OMG framework. How does 
            implementation funding model work in OMG?    (1CIT)
 LeoObrst: Question to EvanWallace: Would a conceptual model -> ontology (logical theory) translation service 
          be desired (e.g., along the line of ODM), and an ontology -> conceptual model translation service, 
          for an OOR?    (1CIU)
 JohnLMcCarthy: Q for BillBug: slide 9 mentions reuse and coordination with other health semantic info 
               efforts; is there any specific relationship to semantics in the National Cancer 
               Institute (NCI) Enterprise Vocabulary Services (EVS)?    (1CIV)
 PeterYim: Question to JohnLMcCarthy: what "services" are available after a certain ontological artifact is 
          "loaded" into XMDR now ... is it consistent across different artifacts?" which I assume the answer 
          is "no." the real question is how are the XMDR folks trying to address that?    (1CIW)
 KevinKeck: @PeterYim: A partial answer is, that the API supports text search and SPARQL query of the 
           registry: https://xmdr.lbl.gov/mediawiki/index.php/Core_APIs    (1CIX)
 KevinKeck: @PeterYim: But in a broader sense, another "service" the registry provides is a metamodel 
           for description of shared and related meanings, both between ontologies and between ontologies 
           and data.    (1CIY)
 JohnLMcCarthy: @PeterYim: to add to Kevin's response, the text and SPARQL query interfaces treat loaded 
               artifacts consistently, in the sense that users can specify queries restricted to particular 
               metadata attributes that are used to characterize the artifacts in a consistent way. When we 
               load metadata (including the individual concept components of ontologies), each component gets 
               mapped into a particular class and attribute of the underlying metamodel.    (1CIZ)
 LeoObrst: Question to JohnLMcCarthy: How come it took so long to load Omega? What format was it in: OWL, etc.? 
          A week: 4 million files, so ~250,000/24 hrs? Is this a limitation of the incoming ontology format 
          or an issue with XMDR? How does this compare, e.g., with various RDF triple stores?    (1CJ0)
 KevinKeck: @LeoObrst: the issue was the engine we've been using to support SPARQL, namely Jena.  We have 
           since mitigated this by using Jena differently (grouping files into much larger models in the 
           Jena store).    (1CJ1)
 MichelleRaymond: For EvanWallace (then other panelists): What are your requirements for (recommendations 
                 for) the ontology repository architecture to best aid in inclusion of support material for 
                 both the stored "Ontology(ies)" and the "Ontology(ies)" inter-relations?   Example in 
                 standards/schema: In working with BPMN, XPDL, BPEL ... (BPMD) - the issues haven't been in 
                 the viability of usage (i.e. schemas can be extended, instances can be generated, they tie 
                 together...). The issues have been lack of explanatory documentation, cookbooks of good 
                 usage examples and libraries, (the things that aid in adoption of the standards.)    (1CJ2)
 LeoObrst: Question to PeterBenson: Given yesterday's story about nuclear missile detonators being 
          delivered to Taiwan, rather than the helicopter batteries they were supposed to be: how 
          can ontologies and an OOR assist in preventing this kind of logistics snafu in the future?    (1CJ3)
 PeterBenson: answer to LeoObrst: I suspect that the problem was a disconnect between the data and the 
             physical object that could be solved by embedding identifiers in objects and resolving the 
             identifiers to metadata from the authoritative source i.e. the manufacturer. This process 
             is used in MSDS where a manufacturer must make available an MSDS for every chemical product 
             they supply.    (1CJ4)
 PeterBenson: correction I meant to say identifiers to master data    (1CJ5)
 RaviSharma: Q for Dr. KenBaclawski. From this varied set of ontologies in allied disciplines, unless 
            we want to overload search engines with SPARQL type Queries, we need to either map namespaces 
            and concepts that are equivalent among ontoloies or look for other efficiencies in IT tools?    (1CJ6)
 KenBaclawski: In response to RaviSharma.  The mappings are often quite complicated.  Species and genus 
              names change over time in many ways, both the classification and the terminology.  There 
              are also several competing taxonomies. One also has many inconsistencies among the ontologies 
              that have to be resolved.    (1CJ7)
 JohnLMcCarthy: Q for RexBrooks: XMDR has tried to represent metadata (including concepts) in a way that 
               conforms to both an XML Schema and a corresponding OWL spec. We wanted to generate one 
               from the other, but that was a challenge. Does ebXML generate one from the other, or are 
               they more loosely linked?    (1CJ8)
 RexBrooks: I wish that were true John. However, the closest I've been able to get is with XMI and that 
           is not especially faithful to either specification.    (1CJ9)
 RexBrooks: We are specifically asking for jurisdictions using EDXL use XMDR to create their own "citable" 
           lists for things like organisational roles, events types, equipment, etc.    (1CJA)
 RexBrooks: What we are planning to do is to make our own specific mappings and publish them along with 
           the specification representations themselves.    (1CJB)
 PeterYim: Q for BillBug, KenBaclawski, SonDoan (and SuziLewis, ChrisChute, MarkMusen ... etc. if they are on this call too) 
          ... how do you see the OOR effort help "normalize" work that among all these related work in 
          the biomedical informatics domain?    (1CJC)
 KenBaclawski: Response to PeterYim. Can you clarify what "normalize" means?    (1CJD)
 RaviSharma: Q for RexBrooks. How do you use decision rules if intended resources are not available? 
            especially for the last slide shown, I can not open it.    (1CJE)
 SonDoan: Q for Dr. KenBaclawski. I has a small question about Environmental ontology (EnvO) and 
         Geo-location instance hierarchy (Gaz). Are they public and how we can access to them ? Thanks.    (1CJF)
 KenBaclawski: Response to SonDoan. Both are public. I will make links available.    (1CJG)
 MichelleRaymond: For BillBug: How do the "users' needs" lead to adding relationship's accross taxonomies?  - 
                 Question based on paraphrased statement, "Given the multiple taxonomies, the relationships 
                 are added - as driven by what the user community expresses as a need.", when speaking on 
                 slide 11 titled 'Objective: represent complex neuroscience domains: Using shared 
                 community ontologies'    (1CJH)
 AnnWrightson: Complementary question to Michelle's, relating to discussion on the [Quality] list: Do 
              participants here have criteria for ontology characteristics that could or should admit 
              or exclude an ontology from an OOR?    (1CJI)
 LeoObrst: Question to all: What is the most important service that an OOR could provide to you, as a 
          content provider? What is the next most important? and the next? ... namely the top 3 services.
          (partially transcribed.)    (1CJJ)
 AnnWrightson: Answer to Leo: a) enabling me to know what ontology-analysis work has already been done 
              in a knowledge-domain (eg in which I have a new project) irrespective of access rules etc; 
              b) characteristics of these existing efforts that enable me to evaluate their potential 
              utility and cost of use/implementation    (1CJK)
 PeterBenson: A to LeoObrst: 1. Protection from claims of "joint work" 2. Mapping 3. Persistance. 
             Taken together beyond unambigious basically data portability and data preservation    (1CJL)
 SonDoan: Answer to LeoObrst. I think the most important service as content provider is how to 
         control the content of ontology. The next is mechanism to allow verify it.    (1CJM)
 EvanWallace (transcribed): Answer to Leo. (1) persistence (availability), (2) maintenance, 
             and (3) support for discovery    (1CJN)
 PeterYim (transacribed): Answer to Leo. ... I guess that will have to be the basics in "open, "ontology" and "repository"
                        (1) that the OOR being *truly* "open",
                        (2) that it does serve *all* sorts of ontological artifacts (not just some, and not others), and
                        (3) that it does provide a *high available* persistent store for those artifacts.    (1CJO)
 KenBaclawski: A to Leo: It is hard to decide which of the requirements is the most important. The ability 
              to browse, query and make inferences across several heterogeneous ontologies efficiently is 
              certainly one of the most important. We can do this now to some degree, but the protocols 
              and formats differ and require ad hoc processing.    (1CJP)
 SonDoan: Additional to Leo. We are developing the BioCaster ontology of infectious disease. There are 
         some experts like epidemiologists, linguists, and anthropologists in US, Japan, Thai, etc involved. 
         For our case it is important to collaborate with others. I think CODS is quite useful. (Thank you 
         very much to PeterYim for hosting our ontology). Also I have a small comment about infrastructure 
         that it would best if it is faster. For our case it takes long time to open ontology.    (1CJQ)
 KenBaclawski: Let me second Son's point. The users lose patience with an ontology based retrieval if it 
         does not have good performance.    (1CJR)
 PeterYim: to Doan -- performance (on the load time for multiuser protege) issue duly noted. In fact, 
          that was a compromise on the part of the protege team to get runtime performance ... I believe 
          some significant improvements is on the way (from the Protege team)    (1CJS)
 Michelle Raymond: For online discussion thread: Given that we are noting as different "levels" 
         - Dictonary, Taxonomy, Ontology, Ontology+, Ontology++, ... is there a different way 
         to manage these within the repository where their "handling" and placement are appropriate 
         for that level? or are all levels just "files" of various quality vectors and should 
         be handled the same?    (1CJT)
 KenBaclawski: To Michelle: While it might be useful for the repository to use different strategies internally, 
              it would not be a good idea for the interface to have diversity. That is one of the problems we 
              now have and that the OOR should help solve.    (1CJU)
 AnnWrightson: Question, relating to discussion on the [Quality] list: Do participants here have criteria 
              for ontology characteristics that could or should admit or exclude an ontology from an OOR?    (1CJV)
 RaviSharma: Response to Peter Benson, there are case based reasoning CBR tools for navigating the FAQs 
            to right professional level of curiosity that can be plugged with search and query tools. 
            This would satisfy and provide answers at appropriate professional levels including the 
            filtering of amateurs' questions at starting levels.    (1CJW)
 LeoObrst: Question to RexBrooks: I am involved in an event management framework effort which is trying to 
          develop a set of ontologies for emergency events. It sounds like EDXL could greatly help in this: 
          where can I get more information? Is there an EDXL ontology or set of ontologies?    (1CJX)
 RexBrooks: Answer to Leo: We are working on this now. As of now we don't have an ontology for this. 
           I can get more information for you by next week.    (1CJY)
 RexBrooks: We are tasking ourselves with developing an EDXL-specific ontology, but we will need to include 
           the work of others: ISCRAM, EIC, DHS, NIEM, DNDO, and several European and Asian communities 
           and organisations as well.    (1CJZ)
 RexBrooks: Perhaps we can discuss this offline, Leo?    (1CK0)
 ToddSchneider (previously anonymous1): I would suggest that 'good' engineering practices should be employed 
               in building ontologies. So that backing evidence or requirements should always be part of 
               the ontology: DOCUMENTATION.    (1CK1)
 ToddSchneider: Another point associated with 'documentation'. It seems to be that many people think that 
               the use of ontologies provides a well-founded (i.e. no infinite descending chains) basis 
               for knowledge representation. I don't see that the use of ontologies provides this 'attribute'. 
               If I'm wrong, please let me know (it's happened before). If not, then rigorous documentation 
               is vital. And documentation should include both references and design decisions (formalized 
               if possible). This sort of information will be/is crucial when it comes time to mediate different 
               terminologies for the same domain.    (1CK2)
 PeterBenson: Who was asking about UDEF?    (1CK3)
 PeterYim: to PeterBenson, it was ThomasBrunner that was asking JohnLMcCarthy about UDEF and ISO 11179, to which 
          both I and FrankOlken responded.    (1CK4)

Audio Recording of this Session    (1C5F)

Join us at part 2 of 2 of this OOR-Panel Discussion - see: ConferenceCall_2008_04_03 !    (1CK5)