[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: Re: [set] Groups - iSURF use case with CAM/SET/OWL dictionary approach(20090121SET-TC-2.ppt) uploaded
Hi Dave, Thank you for your slides; they are very interesting. Correct us if we are wrong but here is what we gathered: You propose to create a Master Dictionary from the UIDs of elements in the document schemas. A part of the Master Dictionary can be constructed automatically by using the CAM Tools, but the rest of the dictionary requires human intervention (i.e. need to be matched manually). By aligning the dictionary of the individual standards with the Master Dictionary will definitely enhance match percentage. In our tools, we use semantics of the document elements and find the relations between them automatically by using the reasoner and rule engine. Please note that we do not include the "Property Terms" of the elements into the semantics that we use, because these terms differ a lot among different standards. However, this lead to some false positives in the discovered matchings. In order to reduce the number of false positivies, we use a string matching algorithm (described in the following paper) between the property terms of the elements: "A String Metric for Ontology Alignment" G Stoilos, G Stamou, S Kollias - LECTURE NOTES IN COMPUTER SCIENCE, 2005 - Springer The algorithm used in the paper finds the similarity between two string terms. We use this similarity value to rank the results found by the reasoner and rule engine, and let the user to select the correct result. As you have also described in your slides, your idea is very useful in this direction. The dictionary approach can be used both to decrease the number of false positives and to enhance the match percentage. Furthermore, we think that when discovering the matchings between the UIDs, you use exact match (Slide number 18). It seems you can use the algorithm described in the above paper to enhance the matchings. Thanks again for your contributions and look forward to hearing your presentation. Best regards, Yildiray & Asuman David RR Webber (XML) wrote: > Greetings from the Starbucks in historic Union Station, Washington > DC. Yes its cold and snowy today in the neighborhood! > > I just posted some ideas for leveraging the CAM template work and UID > dictionary ideas from registry work to enhance the original documents > posted by Asuman earlier this week. > > http://www.oasis-open.org/committees/download.php/30705/20090121SET-TC-2.ppt > > Look forward to discussing this with everyone at our scheduled meeting > and figuring out how we can leverage all this good work together to > make this happen quicker, better and easier. > > I believe the CAM template CXF format is particularly well suited to > use with XSLT to easily garner the information needed to complete the > basic crosswalking. Then with the OWL techniques and other knowledge > tools (reference domain dictionaries) we should be able to provide > enough automation to allow human inspection and arbitration - then > store the results back into the CAM templates for runtime use with > actual transactions. > > Hopefully the PPT slides explain the concepts enough - but obviously > the details needed to be worked into next. > > Thanks, DW > --------------------------------------------------------------------- > To unsubscribe from this mail list, you must leave the OASIS TC that > generates this mail. Follow this link to all your TCs in OASIS at: > https://www.oasis-open.org/apps/org/workgroup/portal/my_workgroups.php -- ____________________________________________________________________________ Professor Asuman Dogac email: asuman@srdc.metu.edu.tr WWW: http://www.srdc.metu.edu.tr/~asuman/ Director Phone: +90 (312) 210 5598, or Software R&D Center +90 (312) 210 2076 Department of Computer Eng. Fax: +90 (312) 210 5572 Middle East Technical University +90 (312) 210 1259 06531 Ankara Turkey skype: adogac
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]