Fatma Arslan

ORCID Identifier(s)


Graduation Semester and Year




Document Type


Degree Name

Doctor of Philosophy in Computer Science


Computer Science and Engineering

First Advisor

Chengkai Li


As social media sites have become major channels for the quick dissemination of news, misinformation has become a significant challenge for our society to tackle. Today fact-checking rests primarily on the shoulders of human fact-checkers who laboriously sift through various trustworthy sources, interview subject experts, and check references before reaching a verdict regarding the degree of truthfulness of a factual claim. Compounded with the speed and scale at which misinformation spreads, the demanding process may leave many harmful factual claims unchecked. In the fight to curb the spread of misinformation, researchers from various disciplines have come forward to assist fact-checkers by creating several automated fact-checking tools and apps. In this dissertation, we focus on studying factual claims and make the following contributions to assist the automated fact-checking efforts: (1) Understanding a factual claim and parsing the content of the claim to extract its attributes are challenging. We propose a way to represent claims in a structured format to capture various aspects of claims, such as entities involved, their relationships, quantities, points and intervals in time, comparisons, and aggregate structures. We use semantic frames for the representation of factual claims. We create a set of new semantic frames, a dataset of frame-annotated claims, and a publicly available web-based annotation tool. (2) To verify a factual claim over a relational database, it is necessary to translate it into a SQL query. However, automatically translating claims to SQL queries is hard. We conduct a preliminary investigative study: (a) to reveal challenges in claim translations and (b) to assess the efficacy of applying a state-of-the-art text-to-SQL parser in translation. (3) The problem of unchecked claims is exacerbated on social media. We build ClaimPortal, a web-based platform that enables users to monitor, search, and check English factual claims on Twitter. We further demonstrate a semantic-frame-based model to categorize tweets based on the type of factual claims they promote. (4) One of the critical components in the fact-checking process is automatically assessing the check-worthiness of a piece of information. It is crucial to have a carefully annotated ground-truth dataset that can feed a machine-learning algorithm to predict the check-worthiness of a statement. To bridge this gap, we create a large dataset of claims from all U.S. presidential debates (1960 to 2016) along with the human-annotated check-worthiness label.


Modeling factual claims, Fact-checking, Frame semantic


Computer Sciences | Physical Sciences and Mathematics


Degree granted by The University of Texas at Arlington