<html xmlns:o="urn:schemas-microsoft-com:office:office" xmlns:w="urn:schemas-microsoft-com:office:word" xmlns:st1="urn:schemas-microsoft-com:office:smarttags" xmlns="http://www.w3.org/TR/REC-html40">
<head>
<meta http-equiv=Content-Type content="text/html; charset=us-ascii">
<meta name=Generator content="Microsoft Word 11 (filtered medium)">
<o:SmartTagType namespaceuri="urn:schemas-microsoft-com:office:smarttags"
name="country-region"/>
<o:SmartTagType namespaceuri="urn:schemas-microsoft-com:office:smarttags"
name="PlaceName"/>
<o:SmartTagType namespaceuri="urn:schemas-microsoft-com:office:smarttags"
name="PlaceType"/>
<o:SmartTagType namespaceuri="urn:schemas-microsoft-com:office:smarttags"
name="City"/>
<o:SmartTagType namespaceuri="urn:schemas-microsoft-com:office:smarttags"
name="place"/>
<!--[if !mso]>
<style>
st1\:*{behavior:url(#default#ieooui) }
</style>
<![endif]-->
<style>
<!--
/* Style Definitions */
p.MsoNormal, li.MsoNormal, div.MsoNormal
{margin:0in;
margin-bottom:.0001pt;
font-size:12.0pt;
font-family:"Times New Roman";}
a:link, span.MsoHyperlink
{color:blue;
text-decoration:underline;}
a:visited, span.MsoHyperlinkFollowed
{color:blue;
text-decoration:underline;}
span.EmailStyle18
{mso-style-type:personal-reply;
font-family:Arial;
color:navy;}
@page Section1
{size:8.5in 11.0in;
margin:1.0in 1.25in 1.0in 1.25in;}
div.Section1
{page:Section1;}
-->
</style>
</head>
<body lang=EN-US link=blue vlink=blue>
<div class=Section1>
<p class=MsoNormal><font size=2 color=navy face=Arial><span style='font-size:
10.0pt;font-family:Arial;color:navy'>Yes, it’s true that the results
depend on the number of eigen vectors. The intuition is that, the more eigen
vectors, the more noise (i.e., accidental co-occurrences) the system
takes into account. With lower numbers of eigen vectors, more
generalization is allowed. <o:p></o:p></span></font></p>
<p class=MsoNormal><font size=2 color=navy face=Arial><span style='font-size:
10.0pt;font-family:Arial;color:navy'><o:p> </o:p></span></font></p>
<p class=MsoNormal><font size=2 color=navy face=Arial><span style='font-size:
10.0pt;font-family:Arial;color:navy'>Ayman has run some tests and found out
that the optimal number of eigen vectors is around 200-300 (this is also true
for the traditional LSA).<o:p></o:p></span></font></p>
<p class=MsoNormal><font size=2 color=navy face=Arial><span style='font-size:
10.0pt;font-family:Arial;color:navy'><o:p> </o:p></span></font></p>
<p class=MsoNormal><font size=2 color=navy face=Arial><span style='font-size:
10.0pt;font-family:Arial;color:navy'>As for the results of the GLSA being
called “similarities”, that’s a topic very much open to
debate indeed. As Peter has said, there is some theoretical motivation to them being
called so (that PMIs can be shown to roughly be equivalent with the original
definition of strengths of association in ACT-R 4.0). Whether association is
the same with similarity, that’s another story. In ACT-R 5.0 I think
there was a tacit understanding (maybe not so tacit, if I remember correctly some
of the discussions at the ACT-R workshops and postgraduate summer school) that
there is some need to experiment with Sji-s and see exactly what they should reflect
(co-occurrence or perhaps some other measure of semantic similarity). <o:p></o:p></span></font></p>
<p class=MsoNormal><font size=2 color=navy face=Arial><span style='font-size:
10.0pt;font-family:Arial;color:navy'><o:p> </o:p></span></font></p>
<p class=MsoNormal><font size=2 color=navy face=Arial><span style='font-size:
10.0pt;font-family:Arial;color:navy'>Raluca<o:p></o:p></span></font></p>
<p class=MsoNormal><font size=2 color=navy face=Arial><span style='font-size:
10.0pt;font-family:Arial;color:navy'><o:p> </o:p></span></font></p>
<p class=MsoNormal><font size=2 color=navy face=Arial><span style='font-size:
10.0pt;font-family:Arial;color:navy'><o:p> </o:p></span></font></p>
<p class=MsoNormal><font size=2 color=navy face=Arial><span style='font-size:
10.0pt;font-family:Arial;color:navy'> <o:p></o:p></span></font></p>
<p class=MsoNormal><font size=2 color=navy face=Arial><span style='font-size:
10.0pt;font-family:Arial;color:navy'><o:p> </o:p></span></font></p>
<div style='border:none;border-left:solid blue 1.5pt;padding:0in 0in 0in 4.0pt'>
<p class=MsoNormal><span class=gmailquote><font size=3 face="Times New Roman"><span
style='font-size:12.0pt'>From: <b><span style='font-weight:bold'>Roman Belavkin</span></b>
<<a href="mailto:R.Belavkin@mdx.ac.uk">R.Belavkin@mdx.ac.uk</a>></span></font></span><br>
<span class=gmailquote>Date: Jun 17, 2005 7:58 AM </span><br>
<span class=gmailquote>Subject: [ACT-R-users] on similarity</span><br>
<span class=gmailquote>To: <a href="mailto:act-r-users@act-r.psy.cmu.edu">act-r-users@act-r.psy.cmu.edu</a></span><br>
<br>
One more observation: The results depend greatly on the number of
eigenvectors used (I understand this controls the number of dimension PCA
reduces the space to). I found the `optimal' value for the word
apple is about 50 dimensions, becuase the results are:<br>
<br>
apple carnation 0.6078935<br>
apple fruit 0.5972236<br>
apple blackberry 0.59559596<br>
apple mac 0.5931967<br>
apple orange 0.570567<br>
apple sweet 0.5677523<br>
apple palm 0.5576204<br>
apple radish 0.55353457<br>
apple intel 0.5511009<br>
apple persimmon 0.54269683<br>
<br>
They are quite different as you can see.<br>
<br>
Cheers!<br>
Roman<br>
<br>
-----Original Message-----<br>
From: <a
href="mailto:act-r-users-admin@act-r.psy.cmu.edu">act-r-users-admin@act-r.psy.cmu.edu</a>
on behalf of Roman Belavkin<br>
Sent: Fri 6/17/2005 15:17<br>
To: Kelley, Troy
(Civ,ARL/HRED); <a href="mailto:act-r-users@act-r.psy.cmu.edu">act-r-users@act-r.psy.cmu.edu
</a><br>
Cc:<br>
Subject: RE: [SPAM: 4.500] RE:
[ACT-R-users] ACT-R output from GLSA server@PARC<br>
<br>
<br>
<br>
Hi,<br>
<br>
I think the word `similar' is
not really appropriate here. LSA just shows co-occurance of the two
words really, and higher co-occurance for word man can be simply explained
because the word man means also human, while woman is more specific and thus
less ambiguous.<br>
<br>
<br>
Cheers,<br>
Roman<br>
<br>
-----Original
Message-----<br>
From:
<a href="mailto:act-r-users-admin@act-r.psy.cmu.edu">act-r-users-admin@act-r.psy.cmu.edu</a>
on behalf of Kelley, Troy (Civ,ARL/HRED)<br>
Sent:
Thu 6/9/2005 21:52<br>
To:
<a href="mailto:act-r-users@act-r.psy.cmu.edu">act-r-users@act-r.psy.cmu.edu</a><br>
Cc:<br>
Subject:
[SPAM: 4.500] RE: [ACT-R-users] ACT-R output from GLSA server@PARC<br>
<br>
<br>
<br>
Here
is a sample output from the GLSA server@PARC for the word Love<br>
<br>
love love 0.9999997<br>
love fun
0.16717705<br>
love city 0.103955254<br>
love place
0.2078263<br>
love man
0.36086833<br>
love friend 0.19896048<br>
love neighbor -0.037641484<br>
love woman
0.21863273<br>
love fondness 0.0076576206<br>
<br>
Interesting,
love is more similar to the word "man" than "woman", and<br>
more
similar to "man" than "fondness" or "friend".<br>
<br>
<st1:City
w:st="on"><st1:place w:st="on">Troy</st1:place></st1:City><br>
<br>
-----Original
Message-----<br>
From:
<a href="mailto:act-r-users-admin@act-r.psy.cmu.edu">act-r-users-admin@act-r.psy.cmu.edu</a><br>
[mailto:<a
href="mailto:act-r-users-admin@act-r.psy.cmu.edu">act-r-users-admin@act-r.psy.cmu.edu</a>]
On Behalf Of <br>
<a
href="mailto:Raluca.Budiu@parc.com">Raluca.Budiu@parc.com</a><br>
Sent:
Thursday, June 09, 2005 3:54 PM<br>
To:
<a href="mailto:act-r-users@act-r.psy.cmu.edu">act-r-users@act-r.psy.cmu.edu</a><br>
Subject:
[ACT-R-users] ACT-R output from GLSA server@PARC<br>
<br>
<br>
<br>
Some
of you may have heard about PARC's effort to build an external<br>
GLSA/PMI
server; it is now available at:<br>
<br>
<a
href="http://glsa.parc.com/">http://glsa.parc.com/</a><br>
<br>
and
it produces ACT-R output (other formats are supported as well).<br>
<br>
GLSA
(Generalized Latent Semantic Analysis) is a LSA-like method of<br>
computing
word similarities, but it has the advantage of an adjustable,<br>
web-based
corpus. It takes as input a list of word pairs and provides<br>
similarities
between those words.<br>
<br>
Just
very recently it started providing ACT-R output; this is an ACT-R<br>
file
that defines a meaning chunk type and sets the Sij-s between words<br>
to
their similarity value as computed by the server.<br>
<br>
The
server is still in a development phase, but please feel free to<br>
experiment
with it. Comments and suggestions are very welcome.<br>
<br>
Raluca
Budiu<br>
<br>
_______________________________________________<br>
ACT-R-users
mailing list<br>
<a href="mailto:ACT-R-users@act-r.psy.cmu.edu">ACT-R-users@act-r.psy.cmu.edu</a><br>
<a
href="http://act-r.psy.cmu.edu/mailman/listinfo/act-r-users">http://act-r.psy.cmu.edu/mailman/listinfo/act-r-users</a>
<br>
<br>
_______________________________________________<br>
ACT-R-users
mailing list<br>
<a
href="mailto:ACT-R-users@act-r.psy.cmu.edu">ACT-R-users@act-r.psy.cmu.edu</a><br>
<a href="http://act-r.psy.cmu.edu/mailman/listinfo/act-r-users">http://act-r.psy.cmu.edu/mailman/listinfo/act-r-users</a><br>
<br>
This
message has been checked for viruses but the contents of an attachment<br>
may
still contain software viruses, which could damage your computer system:<br>
you
are advised to perform your own checks. Email communications with the<br>
<st1:PlaceType
w:st="on">University</st1:PlaceType> of <st1:PlaceName w:st="on">Nottingham</st1:PlaceName>
may be monitored as permitted by <st1:country-region w:st="on"><st1:place
w:st="on">UK</st1:place></st1:country-region> legislation.<br>
<br>
<br>
_______________________________________________<br>
ACT-R-users mailing list<br>
<a
href="mailto:ACT-R-users@act-r.psy.cmu.edu"> ACT-R-users@act-r.psy.cmu.edu</a><br>
<a
href="http://act-r.psy.cmu.edu/mailman/listinfo/act-r-users">http://act-r.psy.cmu.edu/mailman/listinfo/act-r-users</a><br>
<br>
<br>
_______________________________________________ <br>
ACT-R-users mailing list<br>
<a href="mailto:ACT-R-users@act-r.psy.cmu.edu">ACT-R-users@act-r.psy.cmu.edu</a><br>
<a href="http://act-r.psy.cmu.edu/mailman/listinfo/act-r-users">http://act-r.psy.cmu.edu/mailman/listinfo/act-r-users
</a><o:p></o:p></p>
</div>
</div>
</body>
</html>