Doesn't indexing analyzing do this to some degree anyway? Not sure the alogrithm, but something like: How often, hom much near the top, how many differnt forms, subject or object of a sentence. That has to have some relevance to what category something is in.
The simplest extension to that would be something like a 'sub vocabulary' cross listing. If such and such words were hi relevance, then the subject is about this or that. The smartest categorizer is your users, though. So the best way to make that list is to keep track of how close to the top of the search results did a user respond to his search results and what were the words, and how many search attempts did it take. That's waht netflix does. Their goal is to have users get something in theh top three off the first search attempt. Dennis Gearon Signature Warning ---------------- It is always a good idea to learn from your own mistakes. It is usually a better idea to learn from others’ mistakes, so you do not have to make them yourself. from 'http://blogs.techrepublic.com.com/security/?p=4501&tag=nl.e036' EARTH has a Right To Life, otherwise we all die. ----- Original Message ---- From: Erick Erickson <erickerick...@gmail.com> To: solr-user@lucene.apache.org Sent: Thu, December 23, 2010 10:00:05 AM Subject: Re: Item catagorization problem. What you're asking for appears to me to be "auto-categorization", and there's nothing built into Solr to do this. Somehow you need to analyze the documents at index time and add the proper categories, but I have no clue how. This is especially hard with short fields since most auto-categorization algorithms try to do some statistical analysis of the document to figure this out. Best Erick On Thu, Dec 23, 2010 at 8:12 AM, Hasnain <hasn...@hotmail.com> wrote: > > Hi all, > > I am using solr in my web application for search purposes. However, i > am having a problem with the default behaviour of the solr search. > > From my understanding, if i query for a keyword, let's say "Laptop", > preference is given to result rows having more occurences of the search > keyword "Laptop" in the field "name". This, however, is producing > undesirable scenarios, for example: > > 1. I index an item A with "name" value "Sony Laptop". > 2. I index another item B with "name" value: "Laptop bags for laptops". > 3. I search for the keyword "Laptop" > > According to the default behaviour, precedence would be given to item B > since the keyword appears more times in the "name" field for that item. > > In my schema, i have another field by the name of "Category" and, for > example's sake, let's assume that my application supports only two > categories: computers and accessories. Now, what i require is a mechanism > to > assign correct categories to the items during item indexing so that this > field can be used to better filter the search results, item A would belong > to "Computer" category and item B would belong to "Accessories" category. > So > then, searching for "Laptop" would only look for items in the "Computers" > category and return item A only. > > I would like to point out here that setting the category field manually is > not an option since the data might be in the vicinity of thousands of > records. I am not asking for an in-depth algorithm. Just a high level > design > would be sufficient to set me in the right direction. > > thanks. > > > -- > View this message in context: >http://lucene.472066.n3.nabble.com/Item-catagorization-problem-tp2136415p2136415.html >l > Sent from the Solr - User mailing list archive at Nabble.com. >
