Re[2]: daily updates

[Robert Menschel <Ro...@Menschel.net>]

Hello Henry,

Wednesday, August 25, 2004, 9:47:36 AM, you wrote:

HS> If SARE provides me with a corpus and patch file, I will tune and run
HS> the perceptron for them.

Henry, thanks for the offer. We're interested, but each of us uses a
corpus which can include sensitive/confidential material. We'd have to
winnow that out of any corpus we send, which a) lessens the value of the
corpus, and b) takes time we'd rather spend fighting spam.

What we're hoping to do is find a way to emulate the nightly corpus run
used by the development team, such that each night we retrieve whatever
rule sets are to be tested, run mass-check, feed the results of
mass-check back to a central location, and then generate hit-frequencies
and/or perceptron output from that.

Accomplishing that will also mean we can participate in the nightly
corpus run with the development team...

Bob Menschel

Re[2]: daily updates

[Robert Menschel <Ro...@Menschel.net>]

Hello Henry,

Wednesday, August 25, 2004, 7:20:19 PM, you wrote:

HS> Hi Bob,

HS> I should have been more clear.  By "corpus," I meant mass-check
HS> results.  Perhaps we could set it up using the rsync server, the same
HS> way that the nightlies are uploaded.

That would be quite feasible, then.

And actually, we needn't use your resources in the long run -- if you
could help us set up such a system on rulesemporium.com, (whether
initially or after we get it running on your system), we could then do
the rest (doing our SARE perceptrons), and hopefully the experience would
help along the larger daily updates project.

Bob Menschel

Re: daily updates

[Henry Stern <he...@stern.ca>]

Hi Bob,

I should have been more clear.  By "corpus," I meant mass-check 
results.  Perhaps we could set it up using the rsync server, the same 
way that the nightlies are uploaded.

Henry

Robert Menschel wrote:

>Hello Henry,
>
>Wednesday, August 25, 2004, 9:47:36 AM, you wrote:
>
>HS> If SARE provides me with a corpus and patch file, I will tune and run
>HS> the perceptron for them.
>
>Henry, thanks for the offer. We're interested, but each of us uses a
>corpus which can include sensitive/confidential material. We'd have to
>winnow that out of any corpus we send, which a) lessens the value of the
>corpus, and b) takes time we'd rather spend fighting spam.
>
>What we're hoping to do is find a way to emulate the nightly corpus run
>used by the development team, such that each night we retrieve whatever
>rule sets are to be tested, run mass-check, feed the results of
>mass-check back to a central location, and then generate hit-frequencies
>and/or perceptron output from that.
>
>Accomplishing that will also mean we can participate in the nightly
>corpus run with the development team...
>
>Bob Menschel
>
>
>  
>