Home Messages Index
[Date Prev][Date Next][Thread Prev][Thread Next]
Author IndexDate IndexThread Index

Re: Advice on Natural Language Search

__/ [ Jorekunrin@xxxxxxxxx ] on Friday 11 August 2006 01:06 \__

> Hi all,
> 
> I'm developing a location-based search engine for a few New York
> neighborhoods and was hoping that I could get some advice on natural
> language processing. Specifically, I'd like to figure out how to
> answers questions like these:
> 
> 1) "Where can I find the best pizza on the Upper West Side?"
> For questions like this one, I would like the engine to respond by
> searching for Upper West Side restaurants that users have rated highly
> for their pizza.
> 
> 2) "Where do I go to have fun this weekend?"
> With questions like these, I would like the engine to use more
> questions to figure out what the user wants to do to prepare (e.g. "Do
> you want movie recommendations?" "Do you want restaurant
> recommendations?"). Once the user answers a few of these questions, the
> engine could give him specific recommendations. ("Yes? Well, here's
> where this movie is playing in your neighborhood").
> 
> In both cases, the engine has to identify natural language patterns for
> questions. It should also be able to process more technical search
> query methods.
> 
> Anyone who wants more details or would like to give implementation
> suggestions can contact me at jbo25@xxxxxxxxxxxx

Is the group not good enough to accomodate a reply? *disaapointed frown*

Either way, what you describe is something that computer scientists have
attempted to accomplish for a long time. The name of one of the old search
engines, Ask Jeeves, is an example of the ambition. In reality, what you
want to achieve is NP-hard. Indexing of words alone, given the vast scale of
the Web, is one of the greatest problems that human kind is trying to tackle
and spends a lot of resources on (genome and DNA sequencing are among some
other examples).

Either way, you can never read a human mind and it is hard to learn and
traing an engine/knowledgebase when you have incorrect information or
information that needs to be weighed in all sorts of ways (and bad people
will always game the system if it's worthy). The task of parsing natural
language and producing it (synthesis) is pretty much solved, but processing
data of this scale is an impossibility. Even simple theorem provers can only
deal with small chunks of data (they are jamming the comutational clusters
today, too).

Last year, a guy who was not at all interested in the Web (explicitly said
that back in 1993) and said that 64-KB of memory will be enough for everyone
raised courage. He said that Google was lame (still, better than flinging
furniture). He suggested that asking the engine questions and getting
recommendation was the better way. Was he ever to address the issue? On the
face of it -- no. Not only has his venture lost market share, but he also
decided to retire. Maybe there was a reason. Despair maybe? it's a hard
problem. http://justfuckinggoogleit.com/

Best wishes,

Roy
-- 
Roy S. Schestowitz      |    Open Source Othello: http://othellomaster.com
http://Schestowitz.com  |  GNU is Not UNIX  |     PGP-Key: 0x74572E8E
roy      pts/4                         Fri Aug 11 14:28 - 14:31  (00:02)    
      http://iuron.com - proposing a non-profit search engine

[Date Prev][Date Next][Thread Prev][Thread Next]
Author IndexDate IndexThread Index