lucene-solr-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From darniz <rnizamud...@edmunds.com>
Subject Re: Spell check returns strange suggestion
Date Wed, 24 Feb 2010 22:18:53 GMT

Adding on top of this there are business requirement we need to meet.
Let me add we are having a huge content with make, model and their
description. 
Right now we are copying make,model and description to our spelll check
field. 
So our spellcheck fields have words like honda, civic etc.. 
We are also copying description to the spell text field 
So if the description is like this
honda city is a left hand drive then words like hand, drive alos gets
indexed in the spell check index.
The issue is if someone types handa then in our context we would have liked
to get spell suggesetions like
honda instead solr returns back suggestions as hand since hand is a valid
word in our dictionary.

we could have blocked hand in stop words from being indexed to our fiedl but
is this correct way to do things. i mean there might be many more terms
which might resemble to other model names etc..

Does anyone has this kind of issue
Thanks
darniz


darniz wrote:
> 
> Thanks for the prompt reply
> 
> i added the parameter
> <str name="accuracy">0.7</str>
> 
> to my config and this seems to take care of it.
> Works which are present very close to the misspelled words seems to come
> back now.
> 
> darniz
> 
> Markus Jelsma - Buyways B.V. wrote:
>> 
>> darniz said:
>>>
>>> Hello All
>>> Please reply to this ASAP
>>> I am using indexbasedSpellchecker right now i copy only model, and make
>>> names and some other fields to my spellcheck field.
>>> Hence my spell check field consists of only 120 words.
>>>
>>> The issue is if i type hond i get back honda which is fine. But when i
>>> type term like true i get back suggestion like ram.
>> 
>> I'm not quite sure what you're telling us now but you are using
>> onlyMorePopular=true which will almost always return suggestions except
>> if
>> the specified term is actually the most popular.
>> 
>> Another good practice, only show the user suggestions if the
>> correctlySpelled flag is really false. Many users seem to depend on the
>> condition of availability of a collation.
>> 
>> Try turning onlyMorePopular off or rely on the correctlySpelled flag you
>> have.
>> 
>> 
>>>
>>> I read there are some configuration to make for distance measure.
>>> Right now This is my spell check configuration
>>> <lst name="spellchecker">
>>> <str name="name">default</str>
>>> <str name="field">searchSpellText</str>
>>> <str name="buildOnOptimize">true</str>
>>> <str name="buildOnCommit">true</str>
>>> <str
>>> name="distanceMeasure">org.apache.lucene.search.spell.JaroWinklerDistance</str>
>>> <str name="spellcheckIndexDir">./edmundsSpellcheckerDataIndex</str>
>>> </lst>
>>>
>>> and here is my query
>>> q=true&version=2.2&start=0&rows=10&indent=on&spellcheck=true&spellcheck.field=edmundsSearchSpellText&spellcheck.collate=true&spellcheck.extendedResults=true&spellcheck.onlyMorePopular=true
>>>
>>> thanks
>>> darniz
>>> --
>>> View this message in context:
>>> http://old.nabble.com/Spell-check-returns-strange-suggestion-tp27693520p27693520.html
>>> Sent from the Solr - User mailing list archive at Nabble.com.
>> 
>> 
>> 
>> 
>> 
> 
> 

-- 
View this message in context: http://old.nabble.com/Spell-check-returns-strange-suggestion-tp27693520p27714312.html
Sent from the Solr - User mailing list archive at Nabble.com.


Mime
View raw message