From: mailbox@dmitry-kazakov.de   
      
   On Sat, 09 Feb 2008 13:42:08 GMT, calexhc@gmail.com wrote:   
      
   > I am working on a project about book recommendation, I have a   
   > database   
   > contatins basic book's detail .eg.   
   > Title,author,publishdate,price,desc,BookType. When i use my interface   
   > to search on the book, I can get 5 books. And then I want to use   
   > ontology to find the similar book related to a search result. But I   
   > have problem to relate them together.   
   > The problem is: for my ontology, its structure like:   
   > book(root)-->comic-->Action ;"Another path start from book"==>   
   > book(root)-->Fiction-->Horror, For example, If i searched a book "   
   > The Devil Inside ", It has a fuzzy similarity value 0.4 to Action.   
   > And   
   > 0.8 to Horror.(The fuzzy value is defined by me and " The Devil   
   > Inside   
   > "s' type with "Horror and Action" is also defined by me )<==Is that I   
   > have another approach to define the book type rather than defined by   
   > myself? Both the fuzzy value and similar type is predefined in   
   > database.   
      
   Probably you could use clustering. Let you have some primary set of book   
   features describing its genre. Action and Horror can be such, but also they   
   could be rather subsets of this feature space. Clustering would be to   
   determine such subsets in a statistics of data (books in your case)   
   according to some criteria of closeness. Differently to classes you don't   
   know clusters in advance. There are many techniques of fuzzy clustering.   
      
   > Isnt it I can get the result by: 1. from the search result, get the   
   > book's type(Horror and Action) 2. then find just all their subclasses   
   > book of Horror and Action?   
      
   If you could deduce the cluster "Horror and Action" not as an intersection   
   of "Horror" and "Action", but per sole clustering, then that would mean   
   that you already had a measure of distance in the genre feature space. In   
   that case you could just use that distance in order to recommend closest   
   neighbours.   
      
   It would probably try a somewhat mixed approach - a manually designed   
   ontology on top with some clustering in its leaves.   
      
   --   
   Regards,   
   Dmitry A. Kazakov   
   http://www.dmitry-kazakov.de   
      
   [ comp.ai is moderated ... your article may take a while to appear. ]   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)