TY - GEN
T1 - Explicit Stance Detection in the Political Domain
T2 - 17th International Conference on Social Computing, Behavioral-Cultural Modeling and Prediction and Behavior Representation in Modeling and Simulation, SBP-BRiMS 2024
AU - Caceres-Wright, Alexander R.
AU - Udhayasankar, Naveen
AU - Bunn, Grant
AU - Shuster, Stef M.
AU - Joseph, Kenneth
N1 - Publisher Copyright:
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.
PY - 2024
Y1 - 2024
N2 - Stance detection, defined as the task of classifying an individual’s attitude towards a target person or concept, offers the potential to understand political opinions at scale using social media data. However, recent studies have questioned the robustness and accuracy of current stance detection methods, highlighting issues such as generalizability in time and inconsistencies in annotations driven by subtle differences in annotation task design. We argue that central to these challenges is the unresolved question of what constitutes an expression of stance. To address this, the present work introduces a distinction between explicit and implicit stance expressions, and argue that a focus on explicit stance detection addresses many of the existing concerns with modern stance detection methods. To facilitate research on explicit stance detection, we then present a novel (and public) dataset of over 1000 tweets across 13 stance targets for explicit stance detection and evaluate baseline models to establish a foundation for future research in this area.
AB - Stance detection, defined as the task of classifying an individual’s attitude towards a target person or concept, offers the potential to understand political opinions at scale using social media data. However, recent studies have questioned the robustness and accuracy of current stance detection methods, highlighting issues such as generalizability in time and inconsistencies in annotations driven by subtle differences in annotation task design. We argue that central to these challenges is the unresolved question of what constitutes an expression of stance. To address this, the present work introduces a distinction between explicit and implicit stance expressions, and argue that a focus on explicit stance detection addresses many of the existing concerns with modern stance detection methods. To facilitate research on explicit stance detection, we then present a novel (and public) dataset of over 1000 tweets across 13 stance targets for explicit stance detection and evaluate baseline models to establish a foundation for future research in this area.
KW - Large Language Models
KW - Politics
KW - Social Media
KW - Stance Detection
UR - https://www.scopus.com/pages/publications/85205129375
U2 - 10.1007/978-3-031-72241-7_1
DO - 10.1007/978-3-031-72241-7_1
M3 - Conference contribution
AN - SCOPUS:85205129375
SN - 9783031722400
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 3
EP - 14
BT - Social, Cultural, and Behavioral Modeling - 17th International Conference, SBP-BRiMS 2024, Proceedings
A2 - Thomson, Robert
A2 - Pyke, Aryn
A2 - Hariharan, Aravind
A2 - Renshaw, Scott
A2 - Park, Patrick
A2 - Al-khateeb, Samer
A2 - Burger, Annetta
PB - Springer Science and Business Media Deutschland GmbH
Y2 - 18 September 2024 through 20 September 2024
ER -