nl-uiuc AT lists.cs.illinois.edu
Subject: Natural language research announcements
List archive
- From: "Richard Sproat" <rws AT xoba.com>
- To: nl-uiuc AT cs.uiuc.edu
- Subject: [nl-uiuc] Talk on October 4
- Date: Sun, 28 Sep 2008 19:24:43 -0500
- List-archive: <http://lists.cs.uiuc.edu/pipermail/nl-uiuc>
- List-id: Natural language research announcements <nl-uiuc.cs.uiuc.edu>
http://www.acm.uiuc.edu/conference/2008/speakers#RichardSproat
I think you have to register for this thing to attend talks, but registration is free. (See http://www.acm.uiuc.edu/conference/2008/)
11:15 AM
1320 DCL
In a well-known cartoon, Randall Munroe characterizes computational linguistics as a field "so ill-defined that" practitioners "can subscribe to any of dozens of contradictory models and still be taken seriously." As with any humor, there is a serious point here. Indeed, it is often quite hard to see what the substantive claims of computational linguistics are, and how one's model could be falsified (other than by presenting another model that has a lower error rate).
Part of the problem is that a large amount of what counts as computational linguistics is essentially natural language engineering, where the point is simply to find something that works. And part of the problem is a widespread view that has computational linguistics as little more than applied machine learning. Nonetheless, there are models and problems that serve as unifying themes in the field. Among these are the widespread use of finite-state methods, recent work on grammar induction and the growing field of computational modeling of evolutionary language change. Each of these is interesting not only as a computational problem, but also because of the light it may shed on human language.
- [nl-uiuc] Talk on October 4, Richard Sproat, 09/28/2008
Archive powered by MHonArc 2.6.16.