XPost: comp.ai.neural-nets   
   From: kraemert@smail.uni-koeln.de   
      
   Hello fuzzy / neural network implementers   
      
      
   i am building an extension for lucene, the search engine project within   
   apache , that integrates the user´s   
   query history into the retrieval process.   
      
   one idea is to implement a neural network, that adapts to the users   
   information need by reweighting search terms.   
      
   furthermore there are 2 different levels of text representation, one by   
   a large collection of metadata and pure fulltext search, and a second   
   based on the same metadata AND the entire document.   
   the latter is permanently changed in order to obtain a representation   
   closer to what a single user considers relevant.   
      
   first, the application performs a full-text search on a metadata corpus   
   using lucene.   
      
   the user then selects the apparently most interesting documents (by its   
   dc metadata). this second user-input should be used to adapt the neural   
   network weights.   
      
      
   until now, i am not entirely decided which neural network api i should   
   use. joone, at a first glance, seems to be quite user friendly.   
      
   question 1 :   
      
   does an easy to use "term-document-matrix builder" exist, that   
   constructs t-dimensional matrix containing all the terms in the corpus   
   mapped to the documents where they occured?   
   it would be helpful if it were easily possible to compute the   
   tf-idf-measure instead of the absolute term-occurence values (integers).   
      
   i thought of a simple 2-layered nn, in wich one of the layers represents   
   documents and the second one all the terms in the corpus (cf. google:   
   cosimir)   
      
   so, once the users decides to see the full text of several docs   
   refenrenced in metadata retrieved by a full text search according to teh   
   users initial query terms, it must be possible to increase the   
   activation of those neurons that represent the documents selected for   
   full retrieval.   
      
   quetion 2:   
      
   do you know whether there is a more appropriate (open source) software   
   to build such a nn?   
      
   i am a real newbie to ai and appreciate your opinion.   
      
      
   thanks for any hint!   
      
      
   thomas krämer   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)