Fwd: student project on large-scale distributional semantics

[William Cohen (CMU)]

Possible projects (if anyone has plans to switch) on neuroscience -
see below. -W

---------- Forwarded message ----------
From: Brian Murphy <brian...@cmu.edu>
Date: Thu, Mar 22, 2012 at 3:45 PM
Subject: student project on large-scale distributional semantics
To: wco...@cs.cmu.edu
Cc: Alona Fyshe <amf...@gmail.com>

Hello,

I'm looking for willing undergrads interested in helping on one of our
neurosemantics projects, and I understand from Tom that people attending
your large-scale data course might be a good place to start.

We have constructed a range of corpus-semantics models (LSA, HAL, LDA
etc) on the basis of 16 billion words of web-text, which consist of
ten's of thousands of vocabulary words described with 100's of thousands
to millions of features (the large scale bit). And we have several sets
of brain data (fMRI, EEG and MEG; different word sets), with ways of
learning the mapping between word meaning and brain activity. The
question we're trying to answer is what is the optimal corpus-model for
predicting brain states.

Alona told me that your students have already submitted their project
ideas, but in case any of them submit something along these lines, or
their requested project isn't accepted, I'd be happy to talk to them,

thanks,

Brian

--
Brian Murphy
Staff Scientist
Machine Learning Department
Carnegie Mellon University
http://www.cs.cmu.edu/~bmurphy/

--
William W. Cohen
wco...@cs.cmu.edu
http://www.wcohen.com
Research Professor
Machine Learning Department, CMU