[Chicago] Template Document

sheila miguez shekay at pobox.com
Tue Nov 2 20:16:59 CET 2010


On Tue, Nov 2, 2010 at 2:03 PM, Carl Karsten <carl at personnelware.com> wrote:

> for extra credit: The server can use some machine learning foo and use
> previously manually classified stuff as a pattern for new stuff.  like
> if the same phone number comes up a few times, assume that all calls
> to/from that number are part of the same project.

I was at a data mining symposium at depaul the other day where someone
gave a lightning talk on analysing code to trace requirements.

http://dampa.cdm.depaul.edu/events/adms2010jane_huang.html

I never found her to ask after the presentation if they used more than
just code or were throwing all sorts of data at it. I have wanted to
be able to take data such as code, dependencies, repository history
and metadata, bug tracking info, wiki pages, edits, discussions,
forums... everything! and toss it in to a system that could find
clusters of interest that someone could look at and find interesting
things...

like if no one actually knows what the requirements are, you could
have clusters of related data that would maybe fake out some concepts
of what might be 'required' so that when you touch them you know you
might want to talk to so and so product person.

like, if you are wondering what to be paranoid about, you could see
clusters of interest around places that have ever caused problems and
know here be dragons. and know what kind of armor to wear.

oh, and if you want to know who to talk to when you have a question
about a system, this would help identify key people.

there is so much information in all of the data when I start at a
company and usually everything is hidden and inaccessible that you
have to do information archaeology or anthropology to find it.

I submitted this to code.google.com iirc as a feature request. as a
joke, but also a very curious poke because if someone had something
like this, you would imagine it would be them.


-- 
sheila


More information about the Chicago mailing list