Identifier

etd-11152015-141853

Degree

Doctor of Philosophy (PhD)

Department

Engineering Science (Interdepartmental Program)

Document Type

Dissertation

Abstract

In conversations, stories, news reporting, and other forms of natural language, understanding requires participants to make assumptions (hypothesis) based on background knowledge, a process called entailment. These assumptions may then be supported, contradicted, or refined as a conversation or story progresses and additional facts become known and context changes. It is often the case that we do not know an aspect of the story with certainty but rather believe it to be the case; i.e., what we know is associated with uncertainty or ambiguity. In this research a method has been developed to identify different contexts of the input raw text along with specific features of the contexts such as time, location, and objects. The method includes a two-phase SVM classifier along with a voting mechanism in the second phase to identify the contexts. Rule-based algorithms were utilized to extract the context elements. This research also develops a new context˗aware text representation. This representation maintains semantic aspects of sentences, as well as textual contexts and context elements. The method can offer both graph representation and First-Order-Logic representation of the text. This research also extracts a First-Order Logic (FOL) and XML representation of a text or series of texts. The method includes entailment using background knowledge from sources (VerbOcean and WordNet), with resolution of conflicts between extracted clauses, and handling the role of context in resolving uncertain truth.

Date

2015

Document Availability at the Time of Submission

Student has submitted appropriate documentation to restrict access to LSU for 365 days after which the document will be released for worldwide access.

Committee Chair

Knapp, Gerald

DOI

10.31390/gradschool_dissertations.1188

Share

COinS