Dokument: From the Ivory Tower into the Wild - Analysis of (Mis)information in Online Discourses in the Age of Deep Learning
| Titel: | From the Ivory Tower into the Wild - Analysis of (Mis)information in Online Discourses in the Age of Deep Learning | |||||||
| URL für Lesezeichen: | https://docserv.uni-duesseldorf.de/servlets/DocumentServlet?id=66068 | |||||||
| URN (NBN): | urn:nbn:de:hbz:061-20240614-112725-9 | |||||||
| Kollektion: | Dissertationen | |||||||
| Sprache: | Englisch | |||||||
| Dokumententyp: | Wissenschaftliche Abschlussarbeiten » Dissertation | |||||||
| Medientyp: | Text | |||||||
| Autor: | Boland, Katarina [Autor] | |||||||
| Dateien: |
| |||||||
| Beitragende: | Prof. Dr. Dietze, Stefan [Gutachter] Prof. Dr. Mauve, Martin [Gutachter] | |||||||
| Dewey Dezimal-Klassifikation: | 000 Informatik, Informationswissenschaft, allgemeine Werke » 004 Datenverarbeitung; Informatik | |||||||
| Beschreibung: | The Web has evolved into an ubiquitous platform where citizens have the opportunity to express opinions, interact with each other and share information of any kind. In doing so, they leave digital traces, leading to an abundance of Online Discourse Data (ODD). ODD refers to data published online to share opinions, factual claims and other information with other individuals or organizations. It includes messages posted on social media platforms such as Twitter, blog posts, messages in discussion forums, news articles and editorials, scientific information published online, e.g. in digital publications on pre-print servers or in press releases, and any reactions to those such as comments, likes and shares.
ODD has gained relevance as a valuable source of research data for social scientific studies, offering a longitudinal perspective on dynamically evolving debates. Due to the overwhelming amount of information, processing it typically requires the aid of machines. At the same time, ODD has been aiding the advancement of computational methods as it can be used to train powerful, data-intensive Deep Learning models. However, the growing popularity of these methods has been accompanied by concerncs regarding transparency, interpretability and reproducibility of the generated findings and the methods themselves. While they have advanced the state-of-the-art performance in many tasks, scholars are still researching why these methods perform well, and recently, if they really do - or whether their impressive scores are caused by artifacts in the data or evaluation protocols. This uncertainty extends to an imprecise use of terminology within and beyond the research community and inflated expectations of what these methods can achieve, especially since the utility for real-world applications beyond research prototypes has been receiving little attention. This thesis contributes to a clearer understanding of core concepts relating to the analysis of ODD by presenting an extensive multidisciplinary survey and proposing a unifying terminology and model. It further provide insights and tools for the application of state-of-the-art computational methods for Online Discourse Data Analysis in real-world systems to aid fact-checking and combat misinformation. With this, it adds to recent endeavors to assess machine learning methods not merely on the quality of their predictions, but consider their utility for real-world applications. The presented research shows that efficient unupervised systems can be the better choice for verified claim retrieval applications than approaches dominating the leaderboards of respective shared tasks by fine-tuning large language models. Finally, the thesis contributes to the empirical literature on evidence-based policy-making showing that the analysis of ODD can aid in revealing possible interactions between political decisions and public concerns. It further advances the methodological literature on extracting, filtering and analyzing ODD for social scientific studies. | |||||||
| Lizenz: |  Dieses Werk ist lizenziert unter einer Creative Commons Namensnennung 4.0 International Lizenz | |||||||
| Fachbereich / Einrichtung: | Mathematisch- Naturwissenschaftliche Fakultät » WE Informatik | |||||||
| Dokument erstellt am: | 14.06.2024 | |||||||
| Dateien geändert am: | 14.06.2024 | |||||||
| Promotionsantrag am: | 01.06.2023 | |||||||
| Datum der Promotion: | 27.06.2023 |
