One interesting thing to talk about is when you need to create a new Query
subclass, and how to do it.
For example, let's say you want something between a BooleanQuery and a
PhraseQuery, which matches documents with some of the query words in them
(like the normal BooleanQuery), but giving more score to documents which
contain these words near each other (there was a discussion about this idea
about a month ago, when we discussed short documents).
In that case, what do I need to do? I supposed I need to write a new Query
subclass, but what does doing this take? Do I need to write a "Scorer"? A
"Similarity"? Or what?
I think this is an interesting topic.
--
Nadav Har'El
[EMAIL PROTECTED]
+972-4-829-6326
Grant Ingersoll
<[EMAIL PROTECTED]
> To
java-dev@lucene.apache.org
15/06/2006 03:01 cc
AM
Subject
Re: Scoring
Please respond to
[EMAIL PROTECTED]
pache.org
Karl,
This is a great start. I have also started a scoring.xml document under
the xdocs directory (in my sandbox). So far, I have the following
sections (some even have content under them!):
1. Introduction // Intro about Vector Space Model, some references to
theory, links to the Similarity scoring Formula
2. Scoring and the Index //How scoring relates to what is in the index
(i.e. how it takes advantage of precomputed info such as norms, etc.
3. Understanding Similarity //How the Similarity class fits into
Scoring and what it means to override the Similarity (Greek Kung Fu!)
4. Changing Your Scoring -- Expert // A discussion of
overriding/creating Scorer/Query/Whatever else
5. Class Diagrams // Links to your cool pictures
6. Sequence Diagrams //More cool pictures
What else is needed/useful? Anyone want to volunteer on a section?
-Grant
karl wettin wrote:
> On Wed, 2006-06-07 at 08:27 -0400, Grant Ingersoll wrote:
>
>> I have started something in my sandbox that goes in the xdocs directory
>> that is going to cover the scoring and how it works (something parallel
>> in spirit to the file formats documentation). Adding in sequence
>> diagrams and whatever you have would be a perfect fit. I would be happy
>> to coordinate with you, as you may end up getting to it before me.
>>
>> I would also like to see, possibly, some package level documentation and
>> more javadocs.
>>
>
> Day (night) one of me getting to know the finding and scoring of the
> documents matching a query ended up with an initial class diagram.
>
> <
http://wiki.apache.org/jakarta-lucene/KarlWettin?action=AttachFile&do=view&target=search_uml_1.jpg
>
> <http://shorl.com/hynulymolijo>
>
> Feel free to let me know what I got wrong.
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [EMAIL PROTECTED]
> For additional commands, e-mail: [EMAIL PROTECTED]
>
>
>
--
Grant Ingersoll
Sr. Software Engineer
Center for Natural Language Processing
Syracuse University
School of Information Studies
335 Hinds Hall
Syracuse, NY 13244
http://www.cnlp.org
Voice: 315-443-5484
Fax: 315-443-6886
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
