Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!news.duke.edu!concert!sas!mozart.unx.sas.com!saswss
From: saswss@hotellng.unx.sas.com (Warren Sarle)
Subject: Re: CATPAC
Originator: saswss@hotellng.unx.sas.com
Sender: news@unx.sas.com (Noter of Newsworthy Events)
Message-ID: <Cy8tDI.5It@unx.sas.com>
Date: Tue, 25 Oct 1994 19:12:54 GMT
References:  <Cy6wAr.JJs@acsu.buffalo.edu>
Nntp-Posting-Host: hotellng.unx.sas.com
Organization: SAS Institute Inc.
Lines: 31


In article <Cy6wAr.JJs@acsu.buffalo.edu>, comjoew@ubvms.cc.buffalo.edu (Joseph D Woelfel) writes:
|>      Arbitron diaries are simple ASCII text, not numerically encoded. There
|> are hundreds of thousands of them. Before CATPAC, they were read by human
|> beings, who determined the main issues of concern and categorized the
|> responses. This was very expensive.
|>
|>      CATPAC reads the raw text, and assigns a "neuron" in an IAC
|> network to each non-trivial word. It then passes a scanning window through
|> the text; whenever a word is in the window, its corresponding neuron is
|> activated. Synaptic weights are adjusted by a modified Hebb rule, and
|> the resulting square matrix of connection weights is then cluster analyzed
|> and decomposed into a perceptual map, which provides a visual display of
|> the relationshis among the words.

That sounds like great fun, but I don't see how it tells you who listens
to what or whether they like it, which would seem to be the main
questions of interest. Hebbian learning as described above would not
retain information about the identity of particular radio programs or
about which programs are associated with which words. The clustering and
perceptual maps would seem to be providing information about the
respondents semantic behavior, but it is not at all obvious to me (and
clustering and perceptual maps are among my areas of greatest
experience) what this would tell you about the radio programs. So I am
still confused about the point of CATPAC.

-- 

Warren S. Sarle       SAS Institute Inc.   The opinions expressed here
saswss@unx.sas.com    SAS Campus Drive     are mine and not necessarily
(919) 677-8000        Cary, NC 27513, USA  those of SAS Institute.
