Semantic Role Labeler is a machine-learning based tool that analyzes for a shallow semantic information of a given sentence. The tool is capable of outputing verb-argument structure following the notation defined by the Propbank project.
The core of this system is SNoW learning architecture and the Relational Feature Extraction Language (FEX). These core engines were used to learn to analyze a sentence from Propbank section 02-21, specifically the training data provided by CoNLL-2005 shared task.
Constraints provided by experts (e.g., "No two arguments of the same verb can overlap") are modeled as linear inequalities over local decisions, and an Integer Linear Programming system is used to solve for the optimal set of local decisions that satisfy these constraints.
