Mailing-List: contact general-help@lucene.apache.org; run by ezmlm
Precedence: bulk
Reply-To: general@lucene.apache.org
Received-SPF: pass (athena.apache.org: domain of lists@nabble.com designates
 216.139.236.158 as permitted sender)
Message-ID: <17236078.post@talk.nabble.com>
Date: Wed, 14 May 2008 10:02:13 -0700 (PDT)
From: mik07 <someone29_7@yahoo.de>
To: general@lucene.apache.org
Subject: Re: Online Question Answering demo using Lucene
In-Reply-To: <e6537db80805140924v6a86bdbav3a576ad96c089bc0@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
References: <17232494.post@talk.nabble.com>
 <e6537db80805140924v6a86bdbav3a576ad96c089bc0@mail.gmail.com>


Thanks! And you are right, it's roughly the same as Powerset.

It's slower because:
* The demo runs on a single machine (not on a cluster).
* We need to query search engines through their API, which have a 1 second
build-in delay per query.
* We parse sentences once we retrieve them from the search engines and
parsers are still rather slow. Powerset on the other hand, parses Wikipedia
before indexing and indexes the semantic structures. So no parsing needs to
be performed when a user asks a query (beside the parsing of that query, I
suppose.)
* The Lucene index of the complete English Wikipedia we built is 8.3 GB big.
On our machine it takes 2 seconds per query to get a result.

You could address these issues with enough money and man power. But it's
just a research project, developed by one person. We don't have the
resources. (Please drop me an email if you have some ;-)

Cheers,
Michael
-- 
View this message in context: http://www.nabble.com/Online-Question-Answering-demo-using-Lucene-tp17232494p17236078.html
Sent from the Lucene - General mailing list archive at Nabble.com.