<html><body><div style="color:#000; background-color:#fff; font-family:times new roman, new york, times, serif;font-size:12pt"><div><span>Hello all,</span></div><div><span>I am very sorry I did not include xapian-devel mailing list in my previous mail.</span></div><div><span>Thanks for responding my mail.</span></div><div><br><span></span></div><div><span>Mohd Azeem</span></div><div><span>NIT UK</span></div><div style="font-family: times new roman, new york, times, serif; font-size: 12pt;"> <div style="font-family: times new roman, new york, times, serif; font-size: 12pt;"> <div dir="ltr"> <font face="Arial" size="2"> <hr size="1"> <b><span style="font-weight:bold;">From:</span></b> Olly Betts <olly@survex.com><br> <b><span style="font-weight: bold;">To:</span></b> Mohd Azeem <azeem201001@yahoo.in> <br><b><span style="font-weight: bold;">Cc:</span></b> Parth Gupta <parthg.88@gmail.com> <br> <b><span style="font-weight:
bold;">Sent:</span></b> Saturday, 31 March 2012 11:40 AM<br> <b><span style="font-weight: bold;">Subject:</span></b> Re: GSoC, Xapian Project Weighting Schemes<br> </font> </div> <br>Please DON'T mail individual mentors privately - use the xapian-devel<br>mailing list instead.<br><br>On Sat, Mar 31, 2012 at 01:35:16PM +0800, Mohd Azeem wrote:<br>> Presently Xapian<br>> provides the ability to rank search result by the mathematical<br>> formulas like tf*idf andBM25.<br><br>Actually, you can already rank results by incoming hyperlink counts, or<br>any query-independent factor(s) you want to keep track of, and you can<br>combine that with term-based weights. This is done by creating a<br>PostingSource subclass and using it to the query:<br><br><a href="http://xapian.org/docs/postingsource.html" target="_blank">http://xapian.org/docs/postingsource.html</a><br><br>> weight S= S1(Weight calculated by BM25) * S2(weight of document<br>>
calculated based on<br><br>You can't multiply the factors like this with a PostingSource, only add<br>them - is there any theoretical or experimental basis for multiplying<br>the weight contributions in this situation?<br><br>So your suggested project would involve counting up in-bound hyperlinks,<br>and writing a simple PostingSource class to use them, plus perhaps<br>adding a new query operator which multiplies weights. Unfortunately<br>that doesn't seem like it would be nearly enough work for a GSoC<br>project.<br><br>Thanks for the suggestion though.<br><br>Cheers,<br> Olly<br><br><br> </div> </div> </div></body></html>