I'm pretty sure under the algorithm that Chantal describes, if you use a multi-valued field for matching, you're going to get results in your auto-suggest that are in the same document with things that matched your entry, but don't actually match your entry themselves. Chantal seemed to confirm that when I asked. ________________________________________ From: Peter Karich [peat...@yahoo.de] Sent: Friday, September 24, 2010 8:18 AM To: solr-user@lucene.apache.org Subject: Re: Autocomplete: match words anywhere in the token
Jonathan, this field described here from Chantal: > 2.) create an additional field that stores uses the > String type with the same content (use copy field to fill either) can be multivalued. Or what did you mean? BTW: The nice thing about facet.prefix is that you can add an arbitrary (filter) query... Regards, Peter. > Hi Jonathan, > > yes it works only for single-valued fields without great effort. For > multivalued fields you'd have to do some extra work getting only the > values wich contain tokens that start with the given prefix. > > But maybe you mean also wether it works for several fields in one query. > I guess not, but you can create a new field that contains the values of > the fields that you wish to query for autosuggestions (multivalued or > not depending on whether you use facetting or terms comp.). > > I just checked and actually I have such a field, but I use it in > combination with the terms component, while I use the autosuggest based > on facetting in combination with a different single-valued (and > required) field. (I have two different autosuggest sources.) > > However, the suggestions based on the terms component are always single > tokens (because of the way my fields are analyzed) - I haven't put any > effort into changing that because I'm not completely convinced that this > source of suggestions is good in my case here. There are far too many > tokens to suggest from and it all seems very arbitrary. The use case of > autosuggest I have in mind, though, is that of a really long dropdown > box (although of course all entries never show up at once) that offers > complex suggestions (phrases) that really denote some product or person > or other defined objects. And I achieved that with the other > autocomplete based on facets pretty well. > > I definitely need to have a look at how to use facetting in combination > with multivalued fields for autocomplete. > > Cheers, > Chantal > > On Thu, 2010-09-23 at 22:20 +0200, Jonathan Rochkind wrote: > >> This works with _one_ entry per document, right? If you've actually >> found a clever trick to use this technique when you have more than one >> entry for auto-suggest per document, do let me know. Cause I haven't >> been able to come with one. >> >> Jonathan >> >> Chantal Ackermann wrote: >> >>> What works very good for me: >>> >>> 1.) Keep the tokenized field (KeywordTokenizerFilter, >>> WordDelimiterFilter) (like you described you had) >>> 2.) create an additional field that stores uses the String type with the >>> same content (use copy field to fill either) >>> 3.) use facet.prefix instead of terms.prefix for searching the >>> suggestions >>> 4.) to your query add also the String field as a facet, and return the >>> results from that field as suggestion list. They will include the >>> complete String "canon pixma mp500" for example. The other field can >>> only return facets based on tokens. You probably never want that as >>> facets. >>> >>> So your query was alright and the "canon" (2) facet count probably is >>> the two occurrences that you listed, but as the field was tokenized, >>> only tokens would be returned as facets. You need to have an additional >>> field of pure String type to get the complete value as a facet back. >>> >>> In general, it worked out fine for me to create String fields as return >>> values for facets while using the tokenized fields for searching and the >>> actual facet queries. >>> >>> Cheers, >>> Chantal >>> >>> >>> On Wed, 2010-09-22 at 16:39 +0200, Jason Rutherglen wrote: >>> >>> >>>> This may be what you're looking for. >>>> http://www.lucidimagination.com/blog/2009/09/08/auto-suggest-from-popular-queries-using-edgengrams/ >>>> >>>> On Wed, Sep 22, 2010 at 4:41 AM, Arunkumar Ayyavu >>>> <arunkumar.ayy...@gmail.com> wrote: >>>> >>>> >>>>> It's been over a week since I started learning Solr. Now, I'm using the >>>>> electronics store example to explore the autocomplete feature in Solr. >>>>> >>>>> When I send the query terms.fl=name&terms.prefix=canon to terms request >>>>> handler, I get the following response >>>>> <lst name="terms"> >>>>> <lst name="name"> >>>>> <int name="canon">2</int> >>>>> </lst> >>>>> </lst> >>>>> >>>>> But I expect the following results in the response. >>>>> canon pixma mp500 all-in-one photo printer >>>>> canon powershot sd500 >>>>> >>>>> So, I changed the schema for textgen fieldType to use >>>>> KeywordTokenizerFactory and also removed WordDelimiterFilterFactory. That >>>>> gives me the expected result. >>>>> >>>>> Now, I also want the Solr to return "canon pixma mp500 all-in-one photo >>>>> printer" when I send the query terms.fl=name&terms.prefix=pixma. Could >>>>> you >>>>> gurus help me get the expected result? >>>>> >>>>> BTW, I couldn't quite understand the behavior of terms.lower and >>>>> terms.upper >>>>> (I tried these with the electronics store example). Could you also help me >>>>> understand these 2 query fields? >>>>> Thanks. >>>>> >>>>> -- >>>>> Arun >>>>> >>>>> >>>>> >>> >>> >>> > > > -- http://jetwick.com twitter search prototype
