lucene-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Mark Miller <markrmil...@gmail.com>
Subject Re: LUCENE-1124 broken (?)
Date Wed, 14 Oct 2009 14:01:37 GMT
Mark Miller wrote:
> Timo Nentwig wrote:
>   
>> Hi!
>>
>> Consider "abcd" and fuzzy factor 0.75: changing 1 character equals a
>> levensthein distance of exactly 0.75. So isn't it wrong to abandon
>>   term > (1/(1-minSimilarity))
>>
>> ?
>>
>> Wouldn't >= be correct?
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: java-dev-unsubscribe@lucene.apache.org
>> For additional commands, e-mail: java-dev-help@lucene.apache.org
>>
>>     
> I don't think so offhand - but its been a long time since I've had my
> head wrapped around that.
>
> Could you extend the FuzzyQueryTest testTokenLengthOpt() to expose what
> you think the issue is?
>
>   
Just did a quick test myself - if you do a query of abcd and need to
change one char, even if you remove the optimization, it won't match
unless you drop the fuzz below 0.75.

-- 
- Mark

http://www.lucidimagination.com




---------------------------------------------------------------------
To unsubscribe, e-mail: java-dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: java-dev-help@lucene.apache.org


Mime
View raw message