*********************************
There is now a CONTENT FREEZE for Mercury while we switch to a new platform. It began on Friday, March 10 at 6pm and will end on Wednesday, March 15 at noon. No new content can be created during this time, but all material in the system as of the beginning of the freeze will be migrated to the new platform, including users and groups. Functionally the new site is identical to the old one. webteam@gatech.edu
*********************************
Ph.D. Thesis Defense Announcement
Natural language processing and deep learning approaches for
sustainability and infrastructure policy analyses
by
Sooji Ha
Advisor(s):
Dr. Emily Grubert (CEE), Dr. Omar Isaac Asensio (PUBP)
Committee Members:
Dr. Iris Tien (CEE), Dr. Duen Horng Chau (CSE), Dr. Chao Zhang (CSE)
Date & Time: July 7, 2021 | 3:00PM EST
Location: https://bluejeans.com/445422749/3535
Public comments are essential components in the regulatory process, with ability to affect government's rulemaking process. Recently, unsolicited public comments from social media on mobile platforms, have opened new possibilities of engaging the public in government work. Recent computational advances in natural language processing (NLP) and deep learning have shown capabilities for utilizing these types of comments data for policy decision making process. However, when it comes to sustainable infrastructure policy domain, large amounts of publicly available comments data exist but they are yet to fully take advantage of the computational advances. Traditional methods on engaging public opinions for policy implementations include survey and interview processes. However, these survey-based approaches have major limitations as they are often slow and costly to collect. Computationally, unsupervised methods like topic modeling for exploratory analysis of unlabeled texts has been often used. However, such approaches have limitations on creating targeted, theoretically meaningful clusters, and they are not suitable for hypothesis testing, spatial analysis or benchmarking with other corpus.
This dissertation uses NLP with deep learning to overcome these challenges in engaging publicly available comments data with policy making processes. This dissertation expands the literature by demonstrating a framework of designing machine learning lifecycles tailored for sustainable policy analyses. First, I collect and analyzes government solicited public comments towards an energy policy. Next, I assess electric vehicle infrastructure system across the United States using consumer-generated social data using transformer models. Finally, I use attention flow quantification method to interpret the transformer models and examine the model behaviors on predicting the topics of the user generated reviews.
Collectively, this dissertation demonstrates the potential of using public-generated comments and deep learning for research on the sustainable policy analysis and for discovering hard-to-reveal patterns in unstructured large-scale data that can provide useful insights to public policy advisory. The theoretical and methodological contributions of this dissertation help policy makers and industry experts understand the interactions between the public and infrastructure, and make targeted interventions that are effective and equitable.
