MoreLikeThis woes

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

MoreLikeThis woes

Andrew Nagy-2
Hello,  I have been playing off and on with the more like this patch and I really want to get it working well.  I have the patch installed and I have about 500K bibliographic records in my solr index.

My MLT query uses a fieldlist of about 5 or 6 fields.  There are a mix of string and text fields.  They are all in a TermVector.  I have played around with the mindf values.  With about 90% of my mlt queries solr it returns no matches and the remaining 10% get completely irrelevant results.

Are there any "best practices" for the MLT search or am I possibly doing something wrong?

Thanks
Andrew
Reply | Threaded
Open this post in threaded view
|

Re: MoreLikeThis woes

Brian Whitman
> My MLT query uses a fieldlist of about 5 or 6 fields.  There are a  
> mix of string and text fields.  They are all in a TermVector.  I  
> have played around with the mindf values.  With about 90% of my mlt  
> queries solr it returns no matches and the remaining 10% get  
> completely irrelevant results.

Set debugQuery on to see the generated query. Also try mlt.mintf=1 if  
your source fields are singleton-types.






Reply | Threaded
Open this post in threaded view
|

Re: MoreLikeThis woes

Ryan McKinley
In reply to this post by Andrew Nagy-2
Andrew Nagy wrote:
> Hello,  I have been playing off and on with the more like this patch and I really want to get it working well.  I have the patch installed and I have about 500K bibliographic records in my solr index.
>
> My MLT query uses a fieldlist of about 5 or 6 fields.  There are a mix of string and text fields.  They are all in a TermVector.  I have played around with the mindf values.  With about 90% of my mlt queries solr it returns no matches and the remaining 10% get completely irrelevant results.
>
> Are there any "best practices" for the MLT search or am I possibly doing something wrong?
>
> Thanks
> Andrew
>

If you are using MLT on the standard request handler, try debugging with
the MoreLikeThisRequestHandler (/mlt).

as brian suggests, set debugQuery to true and make sure the generated
query looks ok.

also add "&mlt.displayTerms=details" to see what terms are getting used.

I'm not exactly sure what query gets constructed from 6 fields - you may
find some wierdness if different tokenizers/analyzers.  Perhaps try MLT
with 1 or two fields and see if the results make sense, them add more
fields...