Subject: Re: taxonomy discussion at OASIS
Eugene, I agree with the distinction you make between supporting multiple taxonomies and multiple coding schemes. They are indeed quite different. FWIW, ebXML Registry has always had a very flexible classification scheme that supports multi-faceted (i.e. multiple classification schemes) classification using taxonomy trees that may be multi-level deep. At Tokyo we further refined this flexible approach to support context sensitive multi-faceted, multi-level classification. The ebXML Registry specs define a general mechanism for adding such classification or taxonomy trees to the registry while leaving the exact definition of specific taxonomy trees to Core Components team. The net effect is that a registry content may be classified by multiple classification schemes. Each classification is either context free or is defined in a specific context. Please refer to chapter 11 of Registry Information Model v0.41 for details on teh ebXML Registry Classification scheme support. -- Regards, Farrukh Eugene van Roessel wrote: > Duane, > > Good idea. > > I see however a trap in this example. The statement "The registry > allows for multiple classification schemes": > -is true for keyword search, as you showed clearly in your example; > -does not have to be true when browsing through a classification > scheme to find the desired product. To do that, the classification > structure should be stored in the registry. > > The big deal here is that your suggestion promotes using multiple > coding schemes - not taxonomies as I consider it. > > The difference: > - A coding scheme does basically nothing more than attaching a code to > every product. Preferably this is a unique code per product and > preferable the code is unambiguous. > - A taxonomy, or classification scheme, on the other hand defines a > structure between products within that taxonomy. This structure is a > hierarchic structure. > To put it shortly, a taxonomy adds a classification tree, connecting > the products. A coding scheme is merely a set of product codes - > altough the coding might imply a relationship between them. > > So, Duane, I think your suggestion is valuable, but it has to be > recognised that to support a classification scheme fully, the ‘tree > structure above’ should be stored somehow in the registry. > > Eugene > > > At 10:14 22/11/00 -0800, Duane Nickull wrote: > >> Karl: >> >> This subject has been tackled by a few bold individuals. The idea I >> put >> forth was to allow multiple classification schemes within a >> specified >> information model. The RA can allow semantic references suggested >> by >> the SO or Data Steward to be referenced from the repository item. >> An >> Example which may work is a fictional "Skateboard" company. They >> may >> have a regrep item which contains all the information about their >> business including their products. The document instance could >> appear >> similar to this: >> >> <ItemsWeManufacture> >> <item> >> <classification Scheme="NAICS" >> value="SKT4-556">Skateboard</classification> >> <classification Scheme="UNSPSC" >> value="9111234-4444">Skateboard</classification> >> <classification Scheme="UPC-FRANCAIS" value="10764001051">Board du >> Skater</classification> >> </item> >> </ItemsWeManufacture> >> >> This model would allow for efficient retrieval of items by writing a >> >> parser handler that can iterate through the child elements of each >> <item> element seeking for a attribute value equivalent to a >> recognized >> Scheme. A contextual query can also seek for a match on an >> attribute >> value of a recognized scheme and return a hash of positive matches >> in a >> business case where a user was intending to locate manufacturer's of >> a >> specific item. Becuase there can be multiple schemes employed for >> each >> product or item, there are no inherent limitations (hence >> scalability, >> extensibility). >> >> This is also not limited to items for manufacture. >> >> I believe that the OASIS, ebXML and UDDI registry models should >> contain >> information models which allow multiple classification schemes. >> Those >> who wish to re-invent the wheel are free to do so and those who use >> existing standards are also accomodated. >> >> My $0.02 CAD worth (now about $0.013 USD :( ) >> >> Duane Nickull >> >> Karl Best wrote: >> > >> > I've been noticing lately a bit of discussion related to >> taxonomies in the >> > XML e-business space. In particular, most e-business efforts >> require some >> > sort of registry/repository to store information about companies >> with whom >> > an organization may do business with, and the registry needs some >> way to >> > classify the businesses for simple searching. For instance, >> > >> > * The XML.org Registry stores schemas based on keywords, but needs >> a >> > taxonomy for better retrieval, as does the XML.org Portal which is >> seeking >> > to provide XML information sorted by vertical industry. >> > >> > * The ebXML registry/repository will store information about >> businesses, but >> > I'm not sure if a taxonomy has yet been defined. >> > >> > * UDDI will register businesses for the purpose of discovery, and >> has >> > selected for taxonomies NAICS, UNSPSC, an unnamed geographic >> taxonomy, and a >> > general purpose keyword classification (see >> > http://uddi.org/pubs/UDDI_XML_Structure_Reference.pdf page 25). >> > >> > There's a real need to select a good taxonomy (or taxonomies) to >> classify >> > businesses and industries for the work that we're doing in >> creating >> > e-business registries. Notice that I say "select" and not "create" >> -- I >> > don't believe that XML experts should try to reinvent the wheel by >> creating >> > new taxonomies when there are better trained and more experienced >> people in >> > this field who do this stuff for a living. Robin Cover has done a >> bit of >> > research into the subject and has written a background on some of >> the many >> > taxonomies now available; see >> > http://www.oasis-open.org/cover/classification.html. Perhaps one >> or more of >> > these taxonomies would work for what we need. >> > >> > I've set up an email list where we can discuss the use of >> taxonomies in the >> > XML e-business space. I'd like to discussed whether it's possible >> to select >> > an existing taxonomy that would meet the needs of ebXML, UDDI, >> XML.org, >> > etc., what the pros and cons of various taxonomies are, and what >> criteria we >> > should use to select them. The ultimate goal of this list would be >> the >> > creation of an OASIS technical committee, which would select >> taxonomies for >> > use at OASIS and XML.org as well as make recommendations to other >> standards >> > goups. >> > >> > This is an opt-in discussion list: to participate you must >> subscribe to the >> > discussion list by sending a message to >> > taxonomy-discuss-request@lists.oasis-open.org with the word >> "subscribe" as >> > the body of the message. After you are subscribed, you may send >> messages for >> > the list to taxonomy-discuss@lists.oasis-open.org. >> > >> > </karl> >> > ============================================================ >> > Karl F. Best >> > OASIS - Director, Technical Operations >> > 978.667.5115 x206 >> > karl.best@oasis-open.org http://www.oasis-open.org > > > -- Regards, Farrukh
begin:vcard n:Najmi;Farrukh tel;fax:781-442-1610 tel;work:781-442-0703 x-mozilla-html:TRUE url:www.sun.com org:Sun Microsystems;Java Software adr:;;1 Network Drive, MS BUR02-302;Burlington;MA;01803-0902;USA version:2.1 email;internet:najmi@east.sun.com fn:Farrukh S. Najmi end:vcard
Powered by
eList eXpress LLC