Advancing semantic modeling: addressing coordination, interpretability, and data scarcity in domain representation
Year: 2025
Type: dissertation
Abstract: The rapid growth of online text-from scientific articles and product catalogs to news and social media-offers an unprecedented opportunity to uncover structure and meaning from unorganized information.Traditional topic models provide a powerful framework for discovering latent themes but often fall short in practice: their outputs are generic and difficult to interpret, alignment across corpora is not guaranteed, and they struggle with both short documents and domains with limited data.These challenges limit their usefulness for applications such as search, recommendation, trend analysis, and domain comparison, where interpretability and adaptability are essential.This thesis develops four complementary solutions that address these limitations.Coordinated Topic Modeling (CTM) introduces a framework for aligning corpus-specific topics with interpretable reference topics, enabling both interpretability and cross-corpus comparability.Domain Representative Keyword Selection (DRKS) proposes a probabilistic method for extracting distinctive, context-aware keywords that capture the semantics of a target domain.Short-Text Topic Modeling (STTM) leverages large language models and prefix-tuned autoencoders to enrich sparse inputs and produce coherent topics under extreme documentlevel scarcity.Low-Resource Topic Modeling (LRTM) presents DALTA, a domain-adaptation framework that transfers knowledge from data-rich corpora while preserving target-specific nuance.Across multiple datasets and domains, these methods demonstrate substantial improvements over state-of-the-art baselines in coherence, stability, and adaptability.Together, they establish a principled foundation for building structured, interpretable, and adaptable semantic models in diverse and resource-constrained settings.I am deeply grateful for the support and guidance I have received throughout my Ph.D. journey.This dissertation is the result of not only my own efforts but also the encouragement, mentorship, and generosity of many people who have influenced me along the way.First and foremost, I would like to express my heartfelt gratitude to my advisor, Kevin Chen-Chuan Chang.His rigorous research mindset, insightful guidance, and unwavering encouragement have been pivotal in shaping my academic and professional growth.Kevin's ability to balance high expectations with constant support has challenged me to think critically and ambitiously while providing the reassurance needed to navigate setbacks.From shaping the direction of my research to offering thoughtful advice on writing and career development, his mentorship has been the most important factor in my Ph.D. journey.I feel truly privileged to have worked under his supervision, and his influence will continue to guide me well beyond this dissertation.I am also grateful to Professor ChengXiang Zhai, Professor Jingrui He, and Lucian Popa for serving on my dissertation committee and for providing insightful comments, constructive feedback, and valuable suggestions that strengthened this thesis.
Author: Akash, Pritom Saha
Institution: University of Illinois Urbana-Champaign
Language: English
Topic: Topic Modeling
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7220761728
