Date of Award

Spring 1-1-2015

Document Type

Thesis

Degree Name

Master of Arts (MA)

Department

Linguistics

First Advisor

Martha Palmer

Second Advisor

WayneWard

Third Advisor

Mans Hulden

Abstract

PropBank-style (Kingsbury and Palmer, 2002) semantic role labeling has good coverage in several general domains, from the Wall Street Journal Corpus (Palmer et al., 2005) to the medical domain (Albright et al., 2013). The purpose of this project is to explore the efficacy of this labeling schema in the science domain. The My Science Tutor project (Ward et al., 2011) has an abundance of domain-specific data available to evaluate the coverage, portability, and usability of PropBank in sub-domains from the the Full Option Science System (FOSS), such as Energy and Electromagnetism, and Living Systems. The labeler usability will be tested in an off-the-shelf state and compared with manual annotation of the same data, all of which will be taken directly from the My Science Tutor project. A mapping of Propbank- to Phoenix-style annotation will also be devised, and machine learning classifiers for automated output will be created and evaluated.

Included in

Linguistics Commons

Share

COinS