Customize score algorithm

classic Classic list List threaded Threaded
4 messages Options
Reply | Threaded
Open this post in threaded view
|

Customize score algorithm

Diego Bernardes
I'm a long time user of Elasticsearch and Solr, have used in lots of
projects, but until now didn't need to change how the documents are scored.

The actual project my team is developing have a very specific need of
precision and scoring.
Recentely found Algolia, they built a search engine from scratch using a
different aproach from Lucene. The main diffence is on scoring, in my
opinion the way they score is better suited to structured documents (like
products on a ecommerce), the sort is like a sql database, when there are
criterias and they are sorted in order.

The way Lucene works with a single float score works very well for
unestructured documents, but when you know exactly what you want and in
what field, Algolia approach seems better.

The question is, there is any way to do overwrite the way Lucene score? I
need to have several fields or a object, intead of a single float,
something like this:

{
   exactMatch: 2,
   typo: 0,
   customFieldScore: 10
   ...
}

Thanks!
Reply | Threaded
Open this post in threaded view
|

Re: Customize score algorithm

wmartinusa
coir.edu there;s a paper by croft et al about the impact of structured data onslaught being handled by IR engines with strengths in unstructured.

> On Apr 15, 2016, at 1:01 PM, Diego Bernardes <[hidden email]> wrote:
>
> I'm a long time user of Elasticsearch and Solr, have used in lots of
> projects, but until now didn't need to change how the documents are scored.
>
> The actual project my team is developing have a very specific need of
> precision and scoring.
> Recentely found Algolia, they built a search engine from scratch using a
> different aproach from Lucene. The main diffence is on scoring, in my
> opinion the way they score is better suited to structured documents (like
> products on a ecommerce), the sort is like a sql database, when there are
> criterias and they are sorted in order.
>
> The way Lucene works with a single float score works very well for
> unestructured documents, but when you know exactly what you want and in
> what field, Algolia approach seems better.
>
> The question is, there is any way to do overwrite the way Lucene score? I
> need to have several fields or a object, intead of a single float,
> something like this:
>
> {
>   exactMatch: 2,
>   typo: 0,
>   customFieldScore: 10
>   ...
> }
>
> Thanks!

Reply | Threaded
Open this post in threaded view
|

Re: Customize score algorithm

wmartinusa
http://ciir.edu


> On Apr 15, 2016, at 9:53 PM, will martin <[hidden email]> wrote:
>
> coir.edu there;s a paper by croft et al about the impact of structured data onslaught being handled by IR engines with strengths in unstructured.
>
>> On Apr 15, 2016, at 1:01 PM, Diego Bernardes <[hidden email]> wrote:
>>
>> I'm a long time user of Elasticsearch and Solr, have used in lots of
>> projects, but until now didn't need to change how the documents are scored.
>>
>> The actual project my team is developing have a very specific need of
>> precision and scoring.
>> Recentely found Algolia, they built a search engine from scratch using a
>> different aproach from Lucene. The main diffence is on scoring, in my
>> opinion the way they score is better suited to structured documents (like
>> products on a ecommerce), the sort is like a sql database, when there are
>> criterias and they are sorted in order.
>>
>> The way Lucene works with a single float score works very well for
>> unestructured documents, but when you know exactly what you want and in
>> what field, Algolia approach seems better.
>>
>> The question is, there is any way to do overwrite the way Lucene score? I
>> need to have several fields or a object, intead of a single float,
>> something like this:
>>
>> {
>>  exactMatch: 2,
>>  typo: 0,
>>  customFieldScore: 10
>>  ...
>> }
>>
>> Thanks!
>

Reply | Threaded
Open this post in threaded view
|

Re: Customize score algorithm

wmartinusa
crikey.  here it is:   http://ciir.cs.umass.edu/


> On Apr 15, 2016, at 9:54 PM, will martin <[hidden email]> wrote:
>
> http://ciir.edu
>
>
>> On Apr 15, 2016, at 9:53 PM, will martin <[hidden email]> wrote:
>>
>> coir.edu there;s a paper by croft et al about the impact of structured data onslaught being handled by IR engines with strengths in unstructured.
>>
>>> On Apr 15, 2016, at 1:01 PM, Diego Bernardes <[hidden email]> wrote:
>>>
>>> I'm a long time user of Elasticsearch and Solr, have used in lots of
>>> projects, but until now didn't need to change how the documents are scored.
>>>
>>> The actual project my team is developing have a very specific need of
>>> precision and scoring.
>>> Recentely found Algolia, they built a search engine from scratch using a
>>> different aproach from Lucene. The main diffence is on scoring, in my
>>> opinion the way they score is better suited to structured documents (like
>>> products on a ecommerce), the sort is like a sql database, when there are
>>> criterias and they are sorted in order.
>>>
>>> The way Lucene works with a single float score works very well for
>>> unestructured documents, but when you know exactly what you want and in
>>> what field, Algolia approach seems better.
>>>
>>> The question is, there is any way to do overwrite the way Lucene score? I
>>> need to have several fields or a object, intead of a single float,
>>> something like this:
>>>
>>> {
>>> exactMatch: 2,
>>> typo: 0,
>>> customFieldScore: 10
>>> ...
>>> }
>>>
>>> Thanks!
>>
>
http://ciir.cs.umass.edu/