[ACT-R-users] News from GLSA and survey request

[Raluca.Budiu at parc.com]

Hello, all,

A while back I posted on this list advertising the GLSA similarity
computation engine at PARC (http://glsa.parc.com). Thank you all for
your very valuable feedback.  For those of you who haven't checked our
web page in a while, I would like to give you a quick update on the new
status of GLSA project:

1. In the past, because we performed stemming on our corpus, many common
words were reported as not found (e.g., if you entered "places" you
would get an error message because only "place" was actually included in
the database). Now we automatically detect those cases and report the
similarities of the stem words. We soon hope to also have another corpus
(the TASA corpus that LSA uses) available for the similarity
computations.

2. You can use our website to compute raw Pointwise Mutual Information
(PMI) scores and word "frequencies". The  "frequency" of a word
represents in fact the document frequency for that word (i.e., the
number of documents in our corpus that contain that word). The PMI
values can be output into an ACT-R format (as can the GLSA results).

3. If you could spare a few minutes, we would be very grateful if you
could fill in this category rating survey:
http://glsa.parc.com/ratings/

It should take at most 15 minutes. Please do not think too hard about
the answers; there are no correct answers.
We will not be recording any information about who you are, so your
response will be completely anonymous.
This survey is part of our effort to better understand the relationship
between similarity (e.g., GLSA) and category membership.


Thank you in advance and wish you happy holidays!

 --
 Raluca Budiu, Ph.D.
 Palo Alto Research Center
 User Interface Research Group









-- Raluca