|
D. Roth, W. Yih
A Linear Programming Formulation for Global Inference in Natural Language Tasks
AI&Math-04
Special Session on Intelligent Text Processing
The typical processing paradigm in natural language processing is the “pipeline” approach, where learners are being used at one level, their outcomes are being used as features for a second level of predictions and so one. In addition to accumulating errors, it is clear that the sequential processing is a crude approximation to a process in which interactions occur across levels and down stream decisions often interact with previous decisions.
This work develops a general approach to inference over the outcomes of predictors in the presence of general constraints. It allows breaking away from the pipeline paradigm by performing global inference over the outcome of different predictors — potentially learned and evaluated given only partial information — along with domain and task specific constraints on the outcomes of the predictors. At the inference level, the existence of mutual constraints on simultaneous outcomes of predictors results in modifying these predictions to optimize global and task specific constraints.
We develop a linear programming formulation for this problem and evaluate it in the context of simultaneously learning named entities and relations between. Our approach allows us to efficiently incorporate domain and task specific constraints at decision time, resulting in significant improvements in the accuracy and the “human-like” quality of the inferences. @InProceedings{RothYi04,
author = {D. Roth and W. Yih},
title = {A Linear Programming Formulation for Global Inference in Natural Language Tasks},
booktitle = {Proceedings of AI\&Math 2004},
year = {2004}
} |
