Loading ...
Sorry, an error occurred while loading the content.

Re: Faceted Fractals (was...data modelling and taxonomies)

Expand Messages
  • laptopjockey
    Hi Kathleen; I must have missed your post somehow, thank you for the compliment, and apologies to the group if you are getting collectively tired of my
    Message 1 of 11 , Feb 2, 2010
    • 0 Attachment
      Hi Kathleen;

      I must have missed your post somehow, thank you for the compliment, and apologies to the group if you are getting collectively tired of my 'voice'. Apologies or no, these forums are a good way for me to sharpen the ideas I'm working on, so I appreciate the space and the feedback.

      Some of what I'm proposing may seem tedious, but Kathleen's reference to the phrase 'data-driven applications' is a good example of where this is going. Individual data points should be self-describing.

      The two primary issues for me are the almost systemic randomness of classification in the digital universe, and the need for a foundational perspective that supports the richness of the language inside a predictable framework.

      There's a document out there called 'Ontology 101' from Stanford that contains a statement that supports my impression of the first issue. To paraphrase, it says that an ontology is limited in scope by the problem it is trying to solve, and so by definition are unique interpretations of the domain of discourse. You could make the same statement about computer applications, thereby explaining why CRM can't communicate with ERP or CMS. I don't have a problem with this per se, but it does explain a lot.

      Kathleen also mentioned Google as another case in which we are not very well served by the current crop of digital appliances. Indexes - the engines upon which all search is currently built - are dumb. They simply connect a string with a document or a page, and rely on people to add the intelligence.

      The concept of faceted fractals (or quantum semantics, which I call Q6) is an attempt to model the essence of communication as a foundation from which to build all of the rest - however random it may at first appear. There are a number of historical precedents for this: the latin alphabet, the periodic table and the arabic numbering system are all elemental constructs that support infinite combinations.

      So while taxonomy and ontology are absolutely necessary to help organize information I believe they need a foundation from which to connect them to everything else.




      --- In TaxoCoP@yahoogroups.com, "K. McElhinney" <k.mcelhinney@...> wrote:
      >
      > John,
      > Thank you for explaining your ideas in such detail. It is helpful to me, as someone that's relatively new to this world.
      >
      > Reading your email took me back to my student days studying philosophy of mind, where we ran into a similar problem. That time has long been paged out of active memory, so forgive me if some of the following details are a bit hazy.
      >
      > In the philosophy of mind several decades past, one long held belief was finally being put to rest â€" that deductive logic is the model for what goes on in the mind.
      >
      > Like taxonomy, the heart of deductive logic is a binary function â€"the set membership relationship. Either x is in set A or not; there's no half way. But reasoning is rarely that cut and dried. It was the study of vision (in biology, psychology and philosophy) that finally disproved this notion (we perceive change, not newspaper-like dots that we infer images from). Since then, philosophers of mind have pretty much abandoned deductive logic for inductive and statistical explanations.
      >
      > When the field of cognitive science was in its infancy (yes, I'm dating myself), I took a look at what the folks over in electrical engineering were doing to solve similar problems in robotics and AI. One simple suggestion was fuzzy logic â€" which had some success explaining how purposeful arm movement works (e.g., the dozens of minute corrections the brain makes just to pick up a coffee cup). In the intervening decades I lost track of what they were doing, so I don't know if something better has replaced it (maybe Bayesian or statistical?), but fuzzy logic helps me to me understand the difference between taxonomy and ontology, so forgive this short digression.
      >
      > Fuzzy logic uses triples, though not the same way RDF does. In fuzzy logic, there is only one relationship -- 'is a member of'. The third member of the triple is a real number between 0 and 1 (operationally, a percentage or confidence value). This real number is used to indicate to what degree an item is a member of the specified group.
      >
      > I've read about this being used in concept clustering to separate prototypical instances from outliers (e.g., blue jay v. penguin as exemplars for the bird category). If the numbers can be readily assigned, then it works surprisingly well. As I recall, Japanese bullet trains in the 90s used a more complicated version of fuzzy logic to make small course corrections at high speeds.
      >
      > And that was just with one relationship. The semantic web, as Christine defined (below), can be many orders of magnitude more powerful than fuzzy logic. But since most commercial systems are not (yet) 'data-driven applications' (as defined by Bergman; see link below), practioners are still in the dark ages, working with systems not designed to support the relationships we need. I recall the vendor's incredulity when I asked to define my own relationships in a taxonomy management system (about 5 years ago). Though systems can be built that store knowledge in RDF or OWL form, the larger problem seems to be how to find items and then display them coherently in a world fixated on Google. I'm not that familiar with the vendor base, so perhaps others know of systems that solve these problems.
      >
      > To me, the power of ontologies lie in their ability to define and use relationships beyond those in prototypical taxonomies; so I think of ontologies as taxonomies plus 'user-defined' relationships. Thus taxonomies are a small subset of ontologies. Given the realities of time and resources, most of us don't work with ontologies; we work with taxonomies.
      >
      > Since I’m new to ontologies, please let me know if I’m misunderstanding or misstating something.
      >
      > Kathleen McElhinney
      >
      > P.S. My philosophy professors would, of course, point out that nothing we have discussed in this thread is 'really' an ontology â€" describing properties and relationships does not define what a thing IS (its existence). But that's for philosophers to debate. We just need to provide enough context to support the data-driven applications that can provide us with the concrete results our users need. And if the world wants to label it an 'ontology', that's ok with me; it can co-exist in my mind with the philosophical term that uses the same label. :-))
      >
      >
      > --
      > Kathleen McElhinney
      > k.mcelhinney@...
      > 310/833-6073
      > LinkedIn: http://www.linkedin.com/in/kmcelhinney
      >
      > ---- Christine Connors <CJMConnors@...> wrote:
      >
      > =============
      > Hello!
      >
      > These are fascinating comments all, and encompass great ideas and use cases.
      >
      > I would like to ask if ontologies and the linked data web do not meet the criteria set forth?
      >
      > - Clients and servers are flat, but the World Wide Web is n-dimensional, in that any node can be connected to any other node with no cardinality restrictions. (1-to-many).
      > - Triples (subject-predicate-object groupings) are the basic components of an ontological model, just as protons, electrons and neutrons are the basic components of an atom.
      > - These subjects, predicates and objects that make up triples can each
      > have a unique resource identifier on the web, making them nodes that
      > can be connected, 1-to-many.
      > - Triples can be combined into named graphs (lots of triple-based facts linked together) to get a bigger-picture view of information, just as atoms combine to form molecules, molecules form elements, elements form compounds - and so forth as structures get bigger.
      > - Labels for these nodes do not have to be mutually exclusive, as long as the underlying identifier is unique. Therefore, you can have your two "ABC Mechanical" businesses, but each will have different properties defining it in context.
      > - Standard search stinks. Using inferencing engines, reasoners and ontology query tools (e.g. SPARQL) with data visualization techniques from simple table-based structures to interconnected dots in a 3D space provide searchers, analysts and data-archaeologists wonderful discovery capabilities. One example is at <http://bioportal.bioontology.org/visualize/42142#visualization>
      > - When you query across all intelligently mapped models, you can have both improved precision and improved recall.
      > - Researchers and practical application developers have begun building DATA-driven applications. Meaning the application rests on an ontology rather than rigid C or Java or Python (etc) code. See a post from Mike Bergman last May for more. <http://www.mkbergman.com/492/ontology-best-practices-for-data-driven-applications-part-1/>
      >
      > Mike's blog is great, and of course the World Wide Web Consortium is one of the primary agencies behind this vision. <http://www.w3.org/standards/semanticweb/>
      >
      > Thoughts? Comments? Questions? I'd love to hear what taxonomists think of this stuff. :)
      >
      > Best regards,
      > Christine
      >
      > Christine Connors
      > Principal
      > triviumrlg.com
      >
      >
      >
      >
      >
      >
      >
      > ________________________________
      > From: John O'Gorman <jogorman@...>
      > To: TaxoCoP@yahoogroups.com
      > Sent: Mon, January 25, 2010 12:08:35 PM
      > Subject: [TaxoCoP] Faceted Fractals (was...data modelling and taxonomies)
      >
      >
      > Happy New Year, everyone.
      >
      > I hope everyone had a good break and is currently doing something that makes it easy (ok, maybe 'easier') to get out of bed in the morning.
      >
      > I wanted to pick up on the idea that Cherie Wagner lofted into the mix last year; the one about the possibility of fractals being relevant to a new geometry of information. I think it speaks to something that as taxonomists we may tend to miss.
      >
      > There are two things about digital representations of information that everyone who manipulates (organizes and optmizes) digital assets should know about:
      >
      > The first thing is in spite of some very creative configurations to the contrary, the digital universe is flat. Data models, taxonomies, applications, web pages - in fact any hierarchical arrangement of information has historically adopted an approach that mirrors the binary nature of the technology. For example, steps in a taxonomy or folder structure are based on Yes or No decisions.
      >
      > The second thing is the world we live in is multi-dimensional and open to a potentially infinite number of interpretations. The multiple languages, disciplines and media we use to express ourselves are only three facets that support this concept.
      >
      > The net effect of these two facts is the main reason applications are not interoperable. The flatness of the geometry means that in any given implementation an object can only occupy one location at a time. For example, if my objective is to build a taxonomy for my business and one of the nodes is 'Customers' and another is 'Suppliers' where do I put "ABC Mechanical" when they are both? Extend this problem out to an integration exercise between Customer Relationship Management and Enterprise Resource Planning applications and you begin to see where this is headed. ABC Mecahnical can occupy potentially any number of role-based nodes.
      >
      > Another net negative effect is reflected in the way search is set up. Entering 'ABC Mechanical' into a Google appliance may or may not produce the intended results because indexes are another manifestation of the flatness of digital geometry. A search index typically cannot differentiate between 'ABC Mechanical' in Portland and 'ABC Mechanical' in Seattle. Also, the term you are searching for is either referenced in a digital asset or it is not. How many times have you gone searching for a value you know is relevant only to get 'Not Found' back? Or the flipside where you get thousands of hits that you then have to wade through to find what you really want?
      >
      > The new digital geometry says that each string in any given universe of discourse has a sort of signature: a set of orthogonal attributes that persist with that string. This is improtant, because in the new model an entity does not lose its identity when it is used in a new application; so for example 'ABC Mechanical' can take on the application- specific roles of customer, supplier, litigant, etc. without loss of information. Because they are orthogonal (mutually exclusive) the values form an n-dimensional 'shape' that identifies it as unique, while at the same time making it universally recognizable and therefore available for future use. A good working metaphor is chemistry: Carbon has a set of universally recognized properties that at the same time separate it from its neighbours and make it available for use with Hydrogen according to a predictable set of rules.
      >
      > Finishing this post with the example of 'ABC Mechanical', the fractal nature of the new geometry says (among other things) that this string is a Name of an Organization with an affinity (a primary connection) to a Discipline. This particular fractal is a type of "Party" fractal and is very similar to a Person configuration. In the new geometry there are a total of nineteen fractal types just as there are a fixed number of types of elements in the periodic table.
      >
      > As Cherie suggested, collections of fractals of the same type create produce predictible and navigable shapes; while combinations of fractal types can produce recognizable, language-based 'maps'. A taxonomy of your enterprise content - assuming the 'fractalization' of entities - produces a unique arrangement that can be viewed from the perspective of any of the fractal types: information management's take on 3D.
      >
    Your message has been successfully submitted and would be delivered to recipients shortly.