Thanks for your quick response, Seid. There is one more mail I found in the archive[3/4 days old] where someone asked about extracting 3 neighbors words around the match. I think once you have the position of matching term/phrase then extracting 3 or 30 neighbors wont be different, right? because you just have to move back/forward and get the words, this sounds logically simple but I dont know how simple is this implementation-wise. Also people are talking about someting called spanQueries/termvectors etc to use for this purpose. I'm still to get the exact idea of how to do this. As per your mail, you used Java to extract the neighbors, Is that using the standard techniques i.e using those spanqueries/termvectors or something else. If you can elaborate all this a bit It'd be very helpful.
Thank you. KK> On Mon, May 25, 2009 at 10:51 AM, Seid Muhie <seidy...@gmail.com> wrote: > for my thesis work (Question Answering) I used to retrieve first the > document and then play with java to extract the needed answer. > for your case what you will do is first locate the positions of the > query terms in the document (in this case it might be distributed > throughout the document - hence difficult to get the 15/20 words) then > count something10 words forward and backward and extract the match. > this is the way I handle my problem. Hope there might be different I > dea too > > Seid M. > > On 5/25/09, KK <dioxide.softw...@gmail.com> wrote: > > Hi All, > > I'm trying to index some non-english web pages and I'm keeping all the > > content of the page in a single field and the searches are working fine > as > > well. Now when I search for some query it gives the complete page, which > is > > expected. Now I want to restrict the showing of results to say 20 words > > around the match, something like google does, otherwise we cann't make > users > > to look for a match in the whole page content[I'll use highlighter after > > this is done]. So getting positions of the matched word/phrase might help > so > > that I can extract some words before and some words after that and will > show > > that to end user. Any idea on doing the same will be very helpful. Thank > > you. > > > > KK. > > > > > -- > "RABI ZIDNI ILMA" > > --------------------------------------------------------------------- > To unsubscribe, e-mail: java-user-unsubscr...@lucene.apache.org > For additional commands, e-mail: java-user-h...@lucene.apache.org > >
