I know this can be done, because I did it in Ultraseek. It was the first feature I wrote, probably around 1997.
Take the most significant terms from the document and turn them into a query. You can use the tf-idf scores of the terms as weights. I tested my queries (without weights) from Ultraseek against several WWW search engines (we used to have a lot more of those). Two test documents that I remember were the "Yes, Virginia there is a Santa Claus" newspaper column and a Kodak Tri-X film datasheet. For there first, there are lots of copies on the internet. For the second, there are (were) lots of similar 400 speed black and white films. I was an Ilford man, myself. It worked pretty well. wunder On May 14, 2013, at 5:20 PM, Jack Krupansky wrote: > Do you mean pointing to a document in one collection and retrieving documents > from another collection that are similar to the first document? Not that I > know of. > > Sounds like a reasonable Jira request. > > But... the too collections would need to be fairly similar overall > themselves, because the characterization of the first document would be based > on terms in that collection, but those terms might not even occur in the > other collection. > > -- Jack Krupansky > > -----Original Message----- From: SweetEmerald > Sent: Tuesday, May 14, 2013 6:57 PM > To: solr-user@lucene.apache.org > Subject: Search MoreLikeThis ID into another collations > > Hi! > Im new in Solr, I would like to do a MoreLikeThis search for an ID into > other collections, not only in the collection that owns that ID > > Is that posible? do you have an example? > > Thank you very much! > > > > -- > View this message in context: > http://lucene.472066.n3.nabble.com/Search-MoreLikeThis-ID-into-another-collations-tp4063352.html > Sent from the Solr - User mailing list archive at Nabble.com. -- Walter Underwood wun...@wunderwood.org
