Paper: Generalized Inference With Multiple Semantic Role Labeling Systems

ACL ID W05-0625
Title Generalized Inference With Multiple Semantic Role Labeling Systems
Venue International Conference on Computational Natural Language Learning
Session Main Conference
Year 2005
Authors

We present an approach to semantic role labeling (SRL) that takes the output of multiple argument classifiers and com- bines them into a coherent predicate- argument output by solving an optimiza- tion problem. The optimization stage, which is solved via integer linear pro- gramming, takes into account both the rec- ommendation of the classifiers and a set of problem specific constraints, and is thus used both to clean the classification results and to ensure structural integrity of the fi- nal role labeling. We illustrate a signifi- cant improvement in overall SRL perfor- mance through this inference. 1 SRL System Architecture Our SRL system consists of four stages: prun- ing, argument identification, argument classifica- tion, and inference. In particular, the goal of pruning and argumen...