Automatically Acquiring A Semantic Network Of Related Concepts

212 Pages·2016·1.01 MB·English

UUnniivveerrssiittyy ooff CCeennttrraall FFlloorriiddaa
SSTTAARRSS
Electronic Theses and Dissertations, 2004-2019
2013
AAuuttoommaattiiccaallllyy AAccqquuiirriinngg AA SSeemmaannttiicc NNeettwwoorrkk OOff RReellaatteedd
CCoonncceeppttss
Sean Szumlanski
University of Central Florida
Part of the Computer Sciences Commons, and the Engineering Commons
Find similar works at: https://stars.library.ucf.edu/etd
University of Central Florida Libraries http://library.ucf.edu
This Doctoral Dissertation (Open Access) is brought to you for free and open access by STARS. It has been accepted
for inclusion in Electronic Theses and Dissertations, 2004-2019 by an authorized administrator of STARS. For more
information, please contact [email protected].
SSTTAARRSS CCiittaattiioonn
Szumlanski, Sean, "Automatically Acquiring A Semantic Network Of Related Concepts" (2013). Electronic
Theses and Dissertations, 2004-2019. 2585.
https://stars.library.ucf.edu/etd/2585
A A S N
UTOMATICALLY CQUIRING A EMANTIC ETWORK
R C
OF ELATED ONCEPTS
by
S S
EAN ZUMLANSKI
B.S. University of Central Florida, 2004
M.S. University of Central Florida, 2005
A dissertation submitted in partial fulfillment of the requirements
for the degree of Doctor of Philosophy
in the Department of Electrical Engineering and Computer Science
in the College of Engineering and Computer Science
at the University of Central Florida
Orlando, Florida
Spring Term
2013
Major Professor:
Fernando Gomez
© 2013 Sean Szumlanski
ii
ABSTRACT
We describe the automatic acquisition of a semantic network in which over 7,500 of the
most frequently occurring nouns in the English language are linked to their semantically related
concepts in the WordNet noun ontology. Relatedness between nouns is discovered automatically
from lexical co-occurrence in Wikipedia texts using a novel adaptation of an information
theoretic inspired measure. Our algorithm then capitalizes on salient sense clustering among
these semantic associates to automatically disambiguate them to their corresponding WordNet
noun senses (i.e., concepts). The resultant concept-to-concept associations, stemming from 7,593
target nouns, with 17,104 distinct senses among them, constitute a large-scale semantic network
with 208,832 undirected edges between related concepts. Our work can thus be conceived of as
augmenting the WordNet noun ontology with RelatedTo links.
The network, which we refer to as the Szumlanski-Gomez Network (SGN), has been
subjected to a variety of evaluative measures, including manual inspection by human judges and
quantitative comparison to gold standard data for semantic relatedness measurements. We have
also evaluated the network’s performance in an applied setting on a word sense disambiguation
(WSD) task in which the network served as a knowledge source for established graph-based
spreading activation algorithms, and have shown: a) the network is competitive with WordNet
when used as a stand-alone knowledge source for WSD, b) combining our network with
WordNet achieves disambiguation results that exceed the performance of either resource
individually, and c) our network outperforms a similar resource, WordNet++ (Ponzetto & Navigli,
2010), that has been automatically derived from annotations in the Wikipedia corpus.
iii
Finally, we present a study on human perceptions of relatedness. In our study, we elicited
quantitative evaluations of semantic relatedness from human subjects using a variation of the
classical methodology that Rubenstein and Goodenough (1965) employed to investigate human
perceptions of semantic similarity. Judgments from individual subjects in our study exhibit high
average correlation to the elicited relatedness means using leave-one-out sampling (r = 0.77,
σ = 0.09, N = 73), although not as high as average human correlation in previous studies of
similarity judgments, for which Resnik (1995) established an upper bound of r = 0.90 (σ = 0.07,
N = 10). These results suggest that human perceptions of relatedness are less strictly constrained
than evaluations of similarity, and establish a clearer expectation for what constitutes human-like
performance by a computational measure of semantic relatedness. We also contrast the
performance of a variety of similarity and relatedness measures on our dataset to their
performance on similarity norms and introduce our own dataset as a supplementary evaluative
standard for relatedness measures.
iv
That I am able to enjoy a life of safety, liberty, and broad acceptance as an openly gay man is a
privilege that has been purchased for me by the suffering of countless other human beings. This
dissertation is dedicated to those who found the courage to lead open and honest lives in the face
of tremendous adversity, in memory of those who lost their lives for doing so, and to all who
have stood with us in the protracted struggle for LGBT equality.
v
ACKNOWLEDGMENTS
I would first like to thank my committee for their countless contributions to my graduate
career over the years. Charlie Hughes, Annie Wu, and Valerie Sims have been my teachers,
collaborators, co-authors, and mentors. They have impressed me not only with their invaluable
intellectual contributions to my work, but also with how tirelessly they work to set their students
up for success. The encouraging and giving nature of these individuals has made me want to
work harder and give more to my own students and colleagues.
I am particularly grateful to my advisor, Fernando Gomez. He has spent countless hours
in conversation with me over the years, passing on his knowledge not only of artificial
intelligence and computational linguistics, but also of art, philosophy, history, literature, and so
many other things. In my early years under his advisement, one was as likely to find us
discussing Chomsky as one was to find us discussing Lorca and Franco, Picasso’s Guernica,
Bach’s cantatas, or Goya’s witches. I found in him a veritable Abbé Faria (in the Dumasian
sense) whose breadth and depth of knowledge helped make me a more educated and well-
rounded person. My life is richer for our time together, and I am grateful.
Maxine Najle helped facilitate data collection for the perceptions of relatedness study that
is included as part of this dissertation. I owe her huge thanks for that and for giving so
generously of her time during one of the most demanding semesters of her undergraduate career.
I am also grateful to my colleagues from the UCF AI Lab whose discussions, counsel,
distractions, and friendship over the years contributed to my education and personal wellbeing.
In lexicographic order, they are: Adam Campbell, who inspired me with his diligence and
unassuming competence; Adelein Rodriguez, who helped keep me grounded with her
vi
perspectives on life, AI, and so many things in between; Andy Schwartz, who forged ahead of me
on this incredible journey and, in doing so, cleared away some of the thorny brush along the trail
and showed me it was possible to reach the goal; Chris Millward, who impressed me with his
knack for coming up with creative solutions to problems large and small and by being one of the
most level-headed and genuine people I know; Nadeem Mohsin, who made the AI Lab
immeasurably more interesting, geeky, and fun by sharing just some of what is stored in his
amazingly encyclopedic brain; and Ramya Pradhan, who reminded me that sometimes we have
to make sacrifices for the things we want.
I would like to extend my thanks to Arup Guha and Ali Orooji, who have been long-time
teaching mentors and who played critical roles in getting me into the classroom and eventually
having me teach classes on my own. I would also like to thank Ali Orooji and the UCF
Programming Team for welcoming me as a guest at their practices in the fall semesters of 2011
and 2012, where I acquired invaluable experience honing my craft as a programmer and
benefited from their collective knowledge.
I am grateful to the Office of Student Conduct for providing me with so many
opportunities to engage in challenging, meaningful work as a member of UCF’s Student Conduct
Review Board—work that broadened my mind, brought much-needed balance to my life, and
offered me tremendous personal growth. I am particularly grateful to Dana Juntunen, Director of
the Office of Student Rights and Responsibilities, for the special role she played as a mentor to
me during my time on the Board.
I am profoundly grateful to my parents, who have always been there for me, who made
incredible sacrifices to give me a good life, and who told me from a very young age that I could
vii
achieve anything I set my mind to, and to my brother, who is a tremendous friend, whom I
admire for the kindness and compassion in his heart, and who makes me feel understood and
valued for the person I am.
Finally, I would like to thank NASA and the Division of Computer Science at UCF for
their generous financial support over the years. This research was supported in part by the NASA
Engineering and Safety Center under Grant/Cooperative Agreement NNX08AJ98A.
viii
TABLE OF CONTENTS
LIST OF FIGURES......................................................................................................................xiv
LIST OF TABLES........................................................................................................................xvi
LIST OF ACRONYMS AND ABBREVIATIONS........................................................................xx
1 INTRODUCTION....................................................................................................................1
1.1 Semantic Memory and WordNet.....................................................................................2
1.2 Our Contribution.............................................................................................................5
1.3 Corpus Considerations and Co-occurrence.....................................................................8
1.4 Corpus Context and the Limitations of Pre-Specified Relations..................................11
1.5 Using Semantic Resources to Discover Relatedness....................................................13
1.6 Other Parts of Speech....................................................................................................15
1.7 Outline...........................................................................................................................18
1.8 Style Conventions: Words and the Concepts They Denote...........................................20
2 LITERATURE REVIEW........................................................................................................21
2.1 WordNet........................................................................................................................21
2.2 WordNet-Based Measures of Similarity and Relatedness.............................................25
2.2.1 Preliminaries......................................................................................................25
ix