A study on the fundamentals of semantic role labeling

The natural language processing (NLP) community has recently experienced a growing interest in semantic role labeling (SRL) – the process of assigning a WHO did WHAT to WHOM, WHEN, WHERE, WHY and HOW structure to text. The increased availability of annotated resources enables the development of statistical approaches specifically for SRL. This holds potential impact in NLP applications. In this project, we describe the linguistic background of the SRL problem, major resources that are used and an overview of general approaches in computational systems. We reproduce the approaches to SRL based on Pradhan’s ASSERT system extending the work of Gildea and Jurafsky. We examine the system and its individual components, including its annotated resources, parser, classification system, and the features used. We then examine the results obtained by the system and its components. We also assess the challenges in SRL and identify the opportunities for useful further research in SRL.