Otis, I nearly forgot you :)
Attached are the screenshots.
The flow is quite obvious, but as a service here's the architecture
Honestly there's still quite a lot to do, but the basics are already
there. The idea is to be able to handle several corpora per language,
and to be able to have more than one language in the system. Also, we
should be able to remove judgments based on users, and to have an
overall smart system of detecting poor judgments (i.e. keep judgments
from a new user in standby until he judges a few dozens, etc).
There are quite a few questions that come to mind (and some were raised
before), such as what is the ideal way of scoring (boolean, 1..5, other
system), whether we should trust one judgment per doc/topic or we should
try crossing, and so on.
My plan is to give this some attention in a few weeks. As you can see,
its quite a crucial part of my work on Hebrew search. Your thoughts /
On 02/03/2011 04:11, Otis Gospodnetic wrote:
Would you happen to have a screenshot that shows that ORev looks like?
Sematext :: http://sematext.com/
:: Solr - Lucene - Nutch
Lucene ecosystem search :: http://search-lucene.com/
----- Original Message ----
From: Itamar Syn-Hershko<firstname.lastname@example.org>
Sent: Sun, February 27, 2011 2:09:02 PM
Subject: Re: dataset collection
Grant posted a while ago about the ASF mail archive being available from a
cloud store: http://search-lucene.com/m/9udC12n9y5A.
The Orev application (open-relevance viewer) is still under development. I'm
scheduled to resume work on it in a few weeks, and can actually use some help
and feedback. What I have so far is also on github
(https://github.com/synhershko/Orev), and is coded in .NET. I'm planning to
complete this in .NET (and there's still plenty to do, also in terms of design),
unless someone wishes to pick it up and do the actual coding in Java with me
assisting (I'm more fluent with .NET).
On 21/2/2011 6:25 PM, Tommaso Teofili wrote:
I have to use and evaluate a machine learning system for clustering
documents so I am wondering if there is any available dataset used within
ORP I could use.
BTW, is the ORV already in place? May I give you any help with the
development/design of ORP system?