Date of Award

8-25-2011

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Information Science and Systems Engineering

First Advisor

John Talburt

Abstract

In this dissertation, I proposed a novel approach to Named Entity Recognition (NER) in which the contextual and intrinsic indicators are used for locating named entities and their semantic meanings in unstructured textual information (UTI). Named entity is the process of locating a word or a phrase that references a particular entity within a text. The data-intensive approach introduced in this dissertation departs from the traditional Natural Language Processing used in NER tasks in that it does not apply linguistic rules or knowledge in the entity recognition process. It leverages the wide availability of huge amounts of data as well as high-performance computing to provide a NER solution that is independent of the UTI language and subject domain. The use of complex rules, data models and statistical algorithms are substituted with the scanning of annotated large volumes of data at high speeds to find exact or similar known instances of the solution. The proposed NER approach does not require external linguistic resources (e.g. language dictionary) or subject domain resources (e.g. glossary, gazette) in order to acquire the language or subject domain knowledge. Instead, it derives all such knowledge from example documents in which entities of interest have been annotated. The text specific characteristics (language and subject) is decoupled from the analytical processes of locating entities within UTI, consequently, the developed system is agile and can be applied uniformly to UTI in different languages and domains for which annotated example documents exist. A key feature of this NER approach is its ability to determine the semantic meaning or role of an entity within UTI. For example, differentiating between dates of independence and the civil war found in the same document. The motivation behind the proposed NER approach, the different techniques used for locating entities and disambiguating among them, limitations of the system as well as discussion of experiments results are presented in this dissertation.

Download

Share

COinS