"Lukas Vlcek" <[EMAIL PROTECTED]> wrote on 25/10/2007 10:25:23:
> Doron, > > You definitely added few important (crucial) questions. There > are important > concerns and I am glad to hear that Lucene community is > debating them. I am > not an Lucene viscera expert thus I can hardly compare simple > search engine > from the book with Lucene but it is pity to hear that robust > implementation > of the algorithm from the book for Lucene would be very complex, if not > impossible, due to design constraints like non-static docid :-( I am not aware of any search engine implementation whose internal docids are static. Still, please don't let this discussion discourage you - I am not sure that your "needs" from Lucene are sufficiently defined at this stage, and once they are, I'm sure this list will have more useful how-to ideas. > > Back to the customized search, I don't think it will be that > impossible to > store specific neural net for each user speaking about small intranet or > application user base. In the mentioned book only the middle tier (hidden > layer of nodes) from the neural net is kept. Other two layers > of nodes (that > is words in one hand and documents in the other hand) don't have to be > stored in customized database. I believe this is an interesting > idea because > then it is a question of the neural net design (one might need > many hidden > layers while other can live with just very thin net if the results are > sufficient). > > Anyway, thanks for your time. > > Regards, > Lukas You're welcome, Doron --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
