From: Christopher Potts
Date: July 19, 2011 18:36:52 MDT
To: Class list
Subject: Computational Pragmatics: notes and correctives from today

Computational Syntactician/Pragmaticists!

Some quick notes following up on issues today:

1. Chris noticed that there is a problem with the swda-metadata.csv file --- the from caller and to caller info are all the same. I traced this to a bug in my script for pooling all of that information. I've updated the corpus distribution:

http://compprag.christopherpotts.net/code-data/swda.zip

If you've already downloaded it, you can update your version by simply replacing the current metadata file (in the root of your corpus directory) with the one I've attached to this message [direct link: http://compprag.christopherpotts.net/code-data/swda-metadata.csv].  Please do this! The old version is horrible! (A huge thanks to you for catching this, Chris!)

2. I've tried to keep my email list updated with all enrollees and auditors, but the coverage hasn't been great, and people continue to come and go.  To help make amends, I've archived all the course email so far:

http://compprag.christopherpotts.net/email.html

I'll update that with messages to the class as we go. This one is already there!

3. Next time, we'll look at clustering words via dialog-act tag (http://compprag.christopherpotts.net/swda-clustering.html).  I plan to devote the whole meeting to this because the underlying matrix-based perspective on word meanings is powerful and flexible.  If you're especially eager to get started on this, I recommend this paper:

Turney, Peter D. and Pantel, Patrick. 2010. From frequency to meaning: vector space models of semantics. Journal of Artificial Intelligence Research 37: 141-188.
http://www.jair.org/media/2934/live-2934-4846-jair.pdf

A special thanks to everyone who offered new ideas today for clause-typing and other experiments using our new distribution of the SwDA!

---Chris