- From: Kashyap, Vipul <VKASHYAP1@PARTNERS.ORG>
- Date: Tue, 5 Sep 2006 21:09:11 -0400
- To: "William Bug" <William.Bug@DrexelMed.edu>, "Mork, Peter D.S." <pmork@mitre.org>
- Cc: <donald.doherty@brainstage.com>, <public-semweb-lifesci@w3.org>
- Message-ID: <2BF18EC866AF0448816CDB62ADF6538104C16230@PHSXMB11.partners.org>
Sorry for the delay in following up. This kind of discussion is key to getting at the real value proposition of SW specifications. I think one key realization is that SW specifications by themselves are not the answer. It should be accompanied by appropriate architectures (e.g., mediation with a mapping/transformation layer), separation of concerns (e.g., lexical as opposed to semantic), etc. In performing the cross-model mapping, at some point, one needs to specify the semantic relatedness of the attributes in the entities being inter-related. For example, if in one data model you are storing the descriptions about the shape organ sub-regions and descriptions about the shape of cells in the same entity called "biomaterial_shape" with attributes "biomaterial entity" and "shape", while in another source data model these are stored in separate entities called "organ_subregion_shape" and "cell_shape", these two different models need to map in different ways into the "global" mediator schema, and the ways will mainly differ based on the semantic means used to specify different levels of anatomical resolution. [VK] There has been tons of literature in the database field that explores the use of some variant of SQL/relational algebra to represent these mappings. The key question is how can SW specifications like RDF/OWL help? So based on the example given above. We need to specify the following mappings: - DB1.biomaterial_shape.biomaterial_entity <=> DB2.organ_subregion_shape - DB1.biomaterial_shape.shape <=>DB2.cell_shape Please explain why the above mappings are better represented in OWL/RDF as opposed to some variant of relational algebra? How will semantic specifications take care of different levels of anatomical resolution? May be a further drill down on this will help us isolate the value proposition... In some cases (e.g., TAMBIS), the mediating model is actually an ontology, but this is rare. The mappings themselves are expressed in a language with sophisticated transformative capabilities because of the myriad incompatibilities that exist. [BB] Absolutely. These languages are very powerful, but many of the transformational grammars require mixing lexical manipulations and semantic manipulations, as is quite common when using UMLS. This again can introduce inconsistencies in interpretation, as you attempt to build on the first level of semantic correlations you derive from the mappings. [VK] I like this separation of concerns between the "lexical" and the "semantic". The key is to clearly define and identify these type of mappings. For example, aggregation is not supported by many rule or ontology languages. [BB] Again, I completely agree, aggregation is not necessarily supported by some ontological structures, though lexically-oriented graphs are quite good at dealing with aggregations. The Wordnet SynSets are a good example of this. TopicMaps (XTM) also have some advantages to provide in this arena. These derive from the field of text summarization and set theoretical approaches to knowledge organization, which can differ somewhat from the graph theory based methods used in the XML world and often used in ontological formalisms. That's why it can be valuable to keep lexical and semantic formalisms distinct, though the must definitely interoperate. [VK] I am not sure what is meant by "aggregation"? Two simple types of aggregation are: - tuple based aggregation (as in fields of a class definition) - set based aggregation There are ways of expressing these constructs in both SQL and other XML, RDF, OWL types of languages... More complex notions of aggregation, as in "part-of" relationships have not been precisely defined in the first place. However there has been work in FOMA to specify the semantics and appropriate languages (frame based and DLs) have been used to precisely represent these relationships... I guess the issue of aggregation is orthogonal to the notion of semantic and lexical... Neither are regular expressions. [BB]: Regex's are an extremely valuable tool in the lexical analysis toolkit, but they have nothing to do with ontologies. One would hope the results of a Regex-based analysis would be linked to an ontology, but if things are constructed in the proper manner, there is no need for Regex-capability within the ontological framework. [VK] I agree with this characterization completely. The goal of lexical analysis should be to "slot" various data objects into ontological elements and categories. Cheers, ---Vipul ________________________________ From: Kashyap, Vipul [mailto:VKASHYAP1@PARTNERS.ORG] Sent: Tuesday, August 22, 2006 8:57 AM To: William Bug; Mork, Peter D.S. Cc: donald.doherty@brainstage.com; public-semweb-lifesci@w3.org Subject: RE: [HCLS] Bridging Ontology...An Automated Approach? We are using mediation technology within the BIRN project as well. It has many ways in which it can solve some of the problems we're discussing. My fear is every new repository requires an new mapping/registration to the mediator and/or shared ontology. This can be a very fragile system over time, and - as you point out, Peter, tacit assumptions in the source data models (and their use of semantics) - which of course can also change - may greatly limit the depth and breadth with which queries can be mapped to all the data sources. Semantic web approaches appear to circumvent some of the technical frailties, though I don't think they can necessarily overcome the requirement for shared foundational and generic ontology layers. [VK] It will be great if you could give some use cases on how Semantic web approaches can help circumvent the fragility of the system, requiring mappings and registrations... On Aug 22, 2006, at 7:52 AM, Mork, Peter D.S. wrote: Creating explicit connections between all similar and/or identical entries in two schemas is an arduous task that is impractical to do manually. [VK] Will mapping each of these schemas to an ontology and then using the ontology to mediate further queries help alleviate the problem? ---Vipul This is the approach adopted by caBIG. Individual data models (or schemata) are related to a common ontology. By itself, this doesn't allow one to rewrite queries (because of tacit assumptions present in the respective data models?), but the hope is that development of query-mediators will be facilitated by the existence of a loose mapping. Peter Bill Bug Senior Research Analyst/Ontological Engineer Laboratory for Bioimaging & Anatomical Informatics www.neuroterrain.org Department of Neurobiology & Anatomy Drexel University College of Medicine 2900 Queen Lane Philadelphia, PA 19129 215 991 8430 (ph) 610 457 0443 (mobile) 215 843 9367 (fax) Please Note: I now have a new email - William.Bug@DrexelMed.edu <mailto:William.Bug@DrexelMed.edu> This email and any accompanying attachments are confidential. This information is intended solely for the use of the individual to whom it is addressed. Any review, disclosure, copying, distribution, or use of this email communication by others is strictly prohibited. If you are not the intended recipient please notify us immediately by returning this message to the sender and delete all copies. Thank you for your cooperation. Bill Bug Senior Research Analyst/Ontological Engineer Laboratory for Bioimaging & Anatomical Informatics www.neuroterrain.org Department of Neurobiology & Anatomy Drexel University College of Medicine 2900 Queen Lane Philadelphia, PA 19129 215 991 8430 (ph) 610 457 0443 (mobile) 215 843 9367 (fax) Please Note: I now have a new email - William.Bug@DrexelMed.edu This email and any accompanying attachments are confidential. This information is intended solely for the use of the individual to whom it is addressed. Any review, disclosure, copying, distribution, or use of this email communication by others is strictly prohibited. If you are not the intended recipient please notify us immediately by returning this message to the sender and delete all copies. Thank you for your cooperation.
Received on Wednesday, 6 September 2006 01:09:43 UTC