Hi Michael,

it might be possible to make this configurable for fulltext lucene indexes.
>

does it means it is possible or not?

from doc :
http://neo4j.com/docs/milestone/rest-api-indexes.html#rest-api-find-node-by-query
I understood it is possible, by setting :

order=ordering

where ordering is one of index, relevance or score.

http://localhost:7474/db/data/index/node/bobTheIndex?query=Name:Build&order=score

will return the score for each result.
Problem is all the scores are equivalent; as such, it is meaningless - 
sorting will be done anyway as order of id.

>From you previous explanation I understood the issue was in the 
batchimporter.
I tried to re-create the index;

I tried to use constructors from p2neo as well as neo4j-rest-client; I 
think they use constructor
QueryContext
as in http://neo4j.com/docs/milestone/indexing-lucene-extras.html

In chapter "Sorting": We sort the results by relevance (score) like this:

its = movies.query( "title", new QueryContext( "The*" ).sortByScore() );


Is the same as the API rest by setting order=ordering ?


The author of p2neo also suggest to get back to SO and official neo4j about 
full-text indexing.


Summarizing:
I am confused because on one side I understand it is not possible to set 
scoring on lucene indexes due to performance, but it is described in the 
documentation.

I am using neo4j official python drivers, and py2neo seems to support 
lucene indexing with score as documented in neo4j doc:
https://groups.google.com/forum/#!msg/neo4j/9G8fcjVuuLw/adOvqcebuXUJ
Although author suggest to get back to neo4j mailing list or SO for this 
aspect.

If lucene full-text indexes does not work at the moment and is discouraged 
being legacy, which other solutions could be adopted to make a sorted 
full-text query, so to provide paginated results ordered by relevance (so 
that to then select the item and make a traversal)?

Could you suggest any tool or how to proceed?

A comment: if really relevance sorting in full-text indexing is not 
possible due to performance issues and scores in lucene are omitted, I 
think it may be more user-friendly indicate the functionality as 
experimental in the documentation, otherwise one keeps on trying and 
hitting walls without understanding why results are different from ones 
described in documentation.

I do appreciate any constructive suggestion you may already have looked at! 
Thank you so much.


Il giorno domenica 23 agosto 2015 10:40:35 UTC+2, Michael Hunger ha scritto:
>
> Hi,
>
> As explained before, this is due to a performance / size optimization that 
> is currently used which is to omit norms for Neo4j's lucene indexes.
>
> We're looking into that, it might be possible to make this configurable 
> for fulltext lucene indexes.
>
> Michael
>
> Am 20.08.2015 um 17:34 schrieb gg4u <[email protected] <javascript:>>:
>
> I found how to check for the score in :
>
> https://groups.google.com/forum/#!topic/neo4j/9G8fcjVuuLw
>
> and :
>
>
> http://localhost:7474/db/data/index/node/myInde?query=name:myKeyWord&order=score
>
> and I ve found that *ALL THE SCORES ARE SET THE SAME IN LUCENE INDEX*.
>
> here using version* 2.2.1*
>
> *Any help on this?*
>
> I cannot figure out it is a bug at this point.
>
>
>  
>
>
>
>
>
>
>
> Il giorno giovedì 20 agosto 2015 16:58:41 UTC+2, gg4u ha scritto:
>>
>> unfortunately, result hit by the lucene full-text are not yet meaningful: 
>> it is not clear which score or sorting rationale it is used.
>>
>> Is it possible to try out the sort() function of *QueryContext* class in 
>> neo4j-rest-client (or any other python driver)? 
>>
>> something like:
>> i1.query("name","united states").sort()
>>
>> and *then paginate results*?
>>
>> Is it possible to expose the score assigned to a lucene index?
>>
>>
>> http://neo4j.com/docs/stable/indexing-lucene-extras.html#indexing-lucene-sort
>>
>>
>>
>> Il giorno giovedì 20 agosto 2015 14:56:11 UTC+2, gg4u ha scritto:
>>>
>>> I am recreating lucene indexes with neo4j-rest-client:
>>>
>>> http://neo4j-rest-client.readthedocs.org/en/latest/indices.html
>>>
>>>
>>> Will sorting be applied by default by lucene?
>>> If not or not satisfactory, how to change sorting ?
>>>
>>> E.g. syntax in python driver:
>>> i1.add("key", "value", n2)
>>>
>>>  looks like:
>>>
>>> assertContains( index.query( "name", "\"Thomas Anderson\"" ), node );
>>>
>>>
>>> in http://neo4j.com/docs/stable/indexing-create-advanced.html
>>> and we are all happy.
>>>
>>> But I cannot find info about sorting:
>>>
>>> http://neo4j.com/docs/stable/indexing-lucene-extras.html#indexing-lucene-sort
>>>
>>> My goal is to limit the query to a few results, ALREADY sorted
>>> (and not sorting all results having matched a full-text query).
>>>  
>>>
>>> Il giorno mercoledì 19 agosto 2015 23:30:29 UTC+2, gg4u ha scritto:
>>>>
>>>> Hi, I am struggling to overcome the problem that query:
>>>>
>>>> START
>>>>
>>>> against a full-text index will display results not meaningfully sorted.
>>>>
>>>> E.g. using wikipedia as mockup data:
>>>> query 
>>>> 'united states' 
>>>>
>>>> will hit, as first result:
>>>> 'List of United States National Historic Landmarks in United States 
>>>> commonwealths and territories, associated states, and foreign states'
>>>>
>>>>
>>>> I cannot paginate results, cause first results would be meaningless.
>>>> I have to fetch *all *results first, and then (in python) order them - 
>>>> but it takes too long and it s not the way to go.
>>>>
>>>> I posted also a question on SO:
>>>>
>>>> http://stackoverflow.com/questions/31862761/search-queries-in-neo4j-how-to-sort-results-in-neo4j-in-start-query-with-intern
>>>>
>>>> and then found a comment of Michael answering to a similar question:
>>>>
>>>> http://stackoverflow.com/questions/26497068/lucene-in-neo4j-has-some-misbehaviours-in-terms-of-reliable-search-querys-comp
>>>>
>>>>
>>>>> #1 can be handled in Neo4j's Java API by using index.query(new 
>>>>> QueryContext(query).sort(Sort.RELEVANCE));
>>>>
>>>>
>>>> So far I've been learning and using cypher and python, never used java.
>>>> Could you please suggest *any* tutorial or how to in *python*, or at 
>>>> least appoint to which files should I look to modify if JAVA  is the only 
>>>> way, so to obtain a meaningful relevance of results in START query that I 
>>>> can paginate?
>>>>
>>>> I haven't found this aspect in neo4j-rest-client or, if there are, it 
>>>> is not clear to me if sort.relevance is covered:
>>>> http://neo4j-rest-client.readthedocs.org/en/latest/indices.html
>>>>
>>>> It is an important aspect of (my) application cause it allows to start 
>>>> the traversal and any operation on the graph.
>>>>
>>>>
>>>>
> -- 
> You received this message because you are subscribed to the Google Groups 
> "Neo4j" group.
> To unsubscribe from this group and stop receiving emails from it, send an 
> email to [email protected] <javascript:>.
> For more options, visit https://groups.google.com/d/optout.
>
>
>

-- 
You received this message because you are subscribed to the Google Groups 
"Neo4j" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/d/optout.

Reply via email to