Finding the connected elements which make up the neighbourhood is just straightforward lookups of connected IDs on the graph. This can be done using either a database or Lucene - your choice, although I suspect the database is the better choice given the structured nature of the data and any potential volatility in connections. Once you have the IDs which define "the neighbourhood" of nodes you want to search, these IDs can be built into a Lucene filter very fast (see org.apache.lucene.search.TermsFilter.addTerm in the "contrib\queries" section). Using this class I've found Lucene capable of searching neighbourhoods of thousands of nodes very quickly.
The biggest problem you are likely to face is shortlisting the nodes you want to search when "3 degrees of connectivity" leads you at step 1 or 2 to a highly connected node, exploding the list of IDs under consideration. Cheers Mark ----- Original Message ---- From: Sharad Agarwal <[EMAIL PROTECTED]> To: java-user@lucene.apache.org Sent: Monday, 25 September, 2006 10:50:03 AM Subject: searching in social networks I am using lucene for simple flat searches. Now I have a requirement to do searches based on the object's connectivity with other objects. The way the searches are done in "social networks". Lets say I want to search for a query in only those objects which are within 3 degrees of connectivity to a given object. Has any body tried this kind of feature with lucene? Any pointers will be appreciated. thanks sharad --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
