Natural language research announcements

[Julia Hockenmaier <juliahmr AT cs.uiuc.edu>]

Dear all,

this week and next week, we will get a tutorial by Lev Ratinov, Ming- 
Wei Chang and Dan Roth about Constrained Conditional Models for NLP --  
this is a practice for their EACL tutorial, so please come and give  
them feedback!

The schedule for this semester is almost full -- 
http://nlp.cs.uiuc.edu/lunch.html
but if you want to give a talk, please let me know!

Julia


Constrained Conditional Models for NLP

Ming-Wei Chang, Lev Ratinov, Dan Roth

Making decisions in natural language processing problems often  
involves assigning values to sets of interdependent variables where  
the expressive dependency structure can influence, or even dictate,  
what assignments are possible. This setting is of particular  
significance in structured learning problems such as semantic role  
labeling, named entity and relation recognition, co-reference  
resolution, transliteration, summarization and machine translation,  
but the approach has a broader set of applications such as textual  
entailment and question answering. In all these cases, it is natural  
to either formulate the decision process as a constrained optimization  
problem, or to break up the complex problem into a set of subproblems  
and require solutions to be consistent modulo (soft, possibly)  
constraints. In both cases, the resulting objective function is  
composed of learned models, subject to domain or problem specific  
constraints.

Constrained Conditional Models is a learning and inference framework  
that refers to augmenting the learning of conditional (probabilistic  
or discriminative) models with declarative constraints (written, for  
example, using a first-order representation) as a way to support  
decisions in an expressive output space while maintaining modularity  
and tractability of training and inference. Models of this kind have  
recently attracted much attention within the NLP community.

Formulating problems as constrained optimization problems over the  
output of learned models has several advantages. It allows one to  
focus on the modeling of problems by providing the opportunity to  
incorporate problem specific global constraints using a first order  
language, freeing the developer from (much of the) low level feature  
engineering, and it can also guarantee exact inference. It provides  
the freedom of decoupling the stage of model generation (learning)  
from that of the constrained inference stage, often resulting in  
simplifying the learning stage and the engineering problem of building  
an NLP system, while improving the quality of the solutions.

The primary goal of this tutorial is to introduce the framework of  
Constrained Conditional Models (CCMs) to the broader ACL community,  
motivate it as a generic framework for learning and inference in  
global NLP decision problems, present some of the key theoretical and  
practical issues involved in using CCMs and survey some of the  
existing applications of it as a way to promote further development of  
the framework and additional applications. The tutorial will thus be  
useful for many of the senior and junior researchers that have  
interest in global decision problems in NLP, providing a concise  
overview of recent perspectives and research results.