result_similar error

KMandalia · Post by **KMandalia** » Thu Sep 23, 2004 9:04 am

When I enter a query term and click on similar pages link from any record on the first page, the next page has relevance % going as high as 133%. Is this a bug or something is wrong on my part. I just added the <result_similar> function call at the bottom of the result record.

Post by **mark** » Thu Sep 23, 2004 10:25 am

Never seen that before. Try an unmodified search script to see if the same thing happens when you click find similar.

When you say at the "bottom of the result record" where in the script do you mean?

KMandalia · Post by **KMandalia** » Thu Sep 23, 2004 12:19 pm

right next to the url (in similar fashion to what google, yahoo etc. does)

Post by **mark** » Thu Sep 23, 2004 2:09 pm

That's where on the page. I wanted to know where in the script.

Does "find similar" in an unmodified (except for your removal of Keywords of course) script work?

KMandalia · Post by **KMandalia** » Thu Sep 23, 2004 2:30 pm

Yes, it works with original script. I have reverted my changes (No keyword removal in sql)

This is from the original search script:

<a name=result_similar_url export>
<capture><local p><vxcp htmlmode off><$p=$ret>$url$urlq&cmd=mlt&id=$id<vxcp htmlmode $p></capture>
</a>
<a name=result_similar text="Find Similar" class="">
<result_similar_url><a href="$ret"$class>$text</a>
</a>

This is from my modified script:


<a name=result_similar_url export>
<capture><local p><vxcp htmlmode off><$p=$ret>$url$urlq&cmd=mlt&id=$id<vxcp htmlmode $p></capture>
</a>

<a name=result_similar text="Similar Pages" class="">
<result_similar_url><a href="$ret"$class>$text</a>
</a>

I am calling <result_similar> in result2 function.

Post by **mark** » Thu Sep 23, 2004 2:56 pm

I tried that. Can't replicate the errors you're seeing. Is there a url where I can try yours? And maybe show your complete result2 function so I can see how you modified that.

Also, there's no need to modify result_similar. Just put the desired text in the call to it
<result_similar text="Similar Pages">

KMandalia · Post by **KMandalia** » Fri Sep 24, 2004 8:29 am

Also, there's no need to modify result_similar. Just put the desired text in the call to it
<result_similar text="Similar Pages">

--It's working now after I did what you said. Strange!

Thanks.

KMandalia · Post by **KMandalia** » Fri Sep 24, 2004 2:32 pm

http://search.creditunions.com

type in 'indirect lending' and on the first result record, click on 'similar pages'. The no. of results returned is 10 times more and the same record has the relevance of 97%, up from 78%.

While I don't really have a problem with it. I want to know what exactly 'Similar pages' does. Throughout the message board I couldn't find how it works exactly and the manual have a very vague description.

Post by **mark** » Fri Sep 24, 2004 3:59 pm

Similar pages finds pages in the database that have content similar to the one you click on, without respect to the user typed query. Therefore it would be normal to expect the page to be very similar to itself. It finds the similar pages by massaging the text of the page into a query with many terms and performing that search. Note the "texttomm" call in the mlt function.

KMandalia · Post by **KMandalia** » Fri Sep 24, 2004 5:30 pm

I ultimately took it out, since it made no sense whatsoever. Also, the only query word causing the 133% relevance was 'Callahan' (that is the name of our firm, it shows up 1000 times and clicking on similar pages brings 18000 pages). No other term I tried caused the relevance to go beyond 100%. Strange again!