I skim some related research papers, DPH, the parameter free model derived from DfR, achieves good performance.<div>Thus, DPH model should be in. <br clear="all"><div><i><br></i></div><div><i><blockquote class="gmail_quote" style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0.8ex; border-left-width: 1px; border-left-color: rgb(204, 204, 204); border-left-style: solid; padding-left: 1ex; ">
BB2<span class="Apple-tab-span" style="white-space:pre"> </span>Bernoulli-Einstein model with Bernoulli after-effect and normalisation 2.</blockquote><blockquote class="gmail_quote" style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0.8ex; border-left-width: 1px; border-left-color: rgb(204, 204, 204); border-left-style: solid; padding-left: 1ex; ">
IFB2<span class="Apple-tab-span" style="white-space:pre"> </span>Inverse Term Frequency model with Bernoulli after-effect and normalisation 2.</blockquote></i></div><div><i><br></i></div><div><i>These two models don't condition on </i>"classic ad-hoc tasks" or "tasks that require early precision". but I have no idea about their effectiveness.</div>
<div><br></div><div>DFR model also be used for query expansion. I will try to read Enquire::get_eset().</div><div><i><br></i></div><i>Wenjin Wu</i><div><br></div><br>
<br><br><div class="gmail_quote">2011/3/29 Olly Betts <span dir="ltr"><<a href="mailto:olly@survex.com">olly@survex.com</a>></span><br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;">
<div class="im">On Mon, Mar 28, 2011 at 08:26:05PM +0800, wuwenjin wrote:<br>
> As described in <a href="http://terrier.org/docs/current/dfr_description.html" target="_blank">http://terrier.org/docs/current/dfr_description.html</a>, there<br>
> are many DFR models, which models will to be implemented in Xapian.<br>
<br>
</div>DPH is apparently very effective, and it's parameter free (not having<br>
to tune parameters to get the best results is good), so it would<br>
definitely be good to have that one.<br>
<br>
I'm not sure which others are the most interesting. Some models are<br>
better for some situations than others - the page you link to mentions<br>
"classic ad-hoc tasks" and "tasks that require early precision", but I<br>
am not sure which amongst those is the best option.<br>
<br>
I'd suggest picking a representative selection, and aiming to do those.<br>
<br>
There's also scope for implementing DfR query expansion in<br>
Enquire::get_eset() if that interests you. That's probably more<br>
involved since the weights used there aren't pluggable already, so might<br>
be a good "if there's time at the end" thing to look at, once you're<br>
familiar with the query weighting.<br>
<br>
Cheers,<br>
<font color="#888888"> Olly<br>
</font></blockquote></div><br></div>