ORCID Identifier(s)

https://orcid.org/0000-0002-5238-7273

Graduation Semester and Year

Fall 2024

Language

English

Document Type

Dissertation

Degree Name

Doctor of Philosophy in Civil Engineering

Department

Civil Engineering

First Advisor

Kyeong Rok Ryu

Second Advisor

June Young Park

Third Advisor

Mohsen Shahandashti

Fourth Advisor

Joowon Im

Abstract

Infrastructure projects impact a broad range of stakeholders, particularly local communities, whose engagement is critical for successful outcomes. Despite the importance of public engagement in these projects, traditional methods of capturing and analyzing public opinion often fail to fully represent the diverse, genuine perspectives involved. This has led to conflicts between community members and project sponsors. On the other hand, despite advancements in text analytics, including natural language processing (NLP) and its subfields such as topic modeling, sentiment analysis, and neural networks, their functionalities and effectiveness in analyzing public opinion in the domain of infrastructure projects have not been fully investigated.

To address these challenges, this study aims to streamline public engagement in urban transportation infrastructure projects by applying NLP techniques to various communication sources, focusing on the North Houston Highway Improvement Project (NHHIP) case study. Particular emphasis is placed on (1) investigating the spectrum of public opinions on urban infrastructure through online and in-person channels by performing exploratory data analysis on the collected data, (2) conducting a comparative analysis of the performance and effectiveness of sentiment analysis and topic modeling methods across diverse sources and extracting static and temporal insights, and (3) developing a graph attention network for the relation extraction and classification of public concerns and their association with project phases.

The first objective extracts thematic insights and compares results using term-frequency analysis, text summarization, and temporal analysis, revealing an evolving trend in public awareness from general concerns to specific issues as well as the distinct nature of communication channels in representing various public concerns. The second objective evaluates the performance of Vader, Affin, TextBlob, DistilBERT, and RoBERTa for sentiment analysis, and LDA, NMF, and BERTopic for topic modeling on various sources of communication. The results show that traditional lexicon-based methods fail to accurately identify topics and sentiments in public opinion sources. However, RoBERTa excels in sentiment analysis, while BERTopic outperforms traditional methods in topic modeling due to its contextual and semantic capabilities. Finally, a Graph Attention Network (GAT) is developed to identify public concerns and associated project phases, outperforming the Convolutional Neural Network (CNN).

Keywords

Infrastructure, Transportation, Public Engagement, Public Opinion Analysis, Text Analytics, Natural Language Processing, Text Mining, Sentiment Analysis, Topic Modeling, Graph Attention Networks

Disciplines

Civil Engineering | Communication Technology and New Media | Computational Engineering | Computational Linguistics | Computer and Systems Architecture | Construction Engineering and Management | Digital Communications and Networking | Social Justice | Transportation Engineering | Urban Studies and Planning

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.