from:
http://wiki.apache.org/solr/DisMaxQParserPlugin#mm_.28Minimum_.27Should.27_Match.29
"If there are less than 3 optional clauses, they all must match; for 3 to 5
clauses, one less than the number of clauses must match, for 6 or more
clauses, 80% must match, rounded down: "2<-1 5<80%""
Personally, the mm parameter makes my head hurt.
As I read it, there are actually 4 buckets that rules apply to, not three
in your mm definition, see below.
Your mm param says, I think, that
clauses number rule
required
1 1 We haven't gotten to a rule yet, this is the
default
2 2 We haven't gotten to a rule yet, this is the
default
3 2 2<-1
4 3 2<-1
5 2 4<50% rounded down
6 3 5<66% (6 * 0.66 = 3.96)
7 4 5<66% rounded down
Personally, I think the percentages are mind warping and lead to
"interesting" behavior. I prefer to explicitly list the number of causes
required or relatively constant numbers of required clauses, something
like "between 3 and 5, one less. 6 to 9 two less" etc. you don't get
weird steps like between 4 and 5 above. Plus, by the time you get to,
say, 7 clauses nobody can keep track of what correct behavior is anyway <G>.
So I think you're off by one position when applying your rules. Or the Wiki
page is misleading. Or the Wiki page is exactly correct and I'm mis-reading
it.
Like I said, mm makes my head hurt.
Best
Erick
On Fri, Dec 3, 2010 at 8:18 AM, Em <mailformailingli...@yahoo.de> wrote:
>
> Hi list,
>
> I got a little problem with my mm definition:
>
> 2<-1 4<50% 5<66%
>
> Here is what it *should* mean:
>
> If there are 2 clauses, at least one has to match.
> If there are more than 2 clauses, at least 50% should match (both rules
> seem
> to mean the same, don't they?).
> And if there are 5 or more than 5 claues, at least 66% should match.
>
> In case of 5 clauses, 3 should match, in case of 6 at least 4 should match
> and so on.
>
> However in some test-case I get only the intended behaviour with a
> 2-clause-query when I say mm=1.
> If I got longer queries this would lead to very bad search-quality-results.
>
> What is wrong with this mm-definition?
>
> Thanks for suggestions.
> - Em
> --
> View this message in context:
> http://lucene.472066.n3.nabble.com/Problem-with-dismax-mm-tp2011496p2011496.html
> Sent from the Solr - User mailing list archive at Nabble.com.
>
