Tools
Tools
Dasika, Vijaya Krishna Likhith (2024) Advanced Fake News Detection Using ALBERT and Knowledge Graphs. Masters thesis, Dublin, National College of Ireland.
Preview |
PDF (Master of Science)
Download (615kB) | Preview |
Preview |
PDF (Configuration Manual)
Download (491kB) | Preview |
Abstract
Because fake news has posted serious problems to information integrity, public discourse, and digital media has suffered the most, this paper proposes a new approach in detection using advanced language models with knowledge representation. In this paper, ALBERT is used together with Knowledge Graphs for capturing complex patterns at the linguistic and contextual levels of news articles. A dataset of 10,395 labeled news items has been pre-processed, ALBERT-encoded, and the knowledge graph constructed. Results in a hybrid model of good performance: an accuracy of 70.19% and an AUC of 0.7026 are achieved. More compelling is its nuanced detection ability, it had attained a recall of 77% for fake news identification against 63% for genuine articles. This research illustrates the combination of advanced natural language processing techniques with structured knowledge frameworks that enormously improve detections of false information. This described approach could serve as a good basis for coming up with much more advanced and complex systems to withstand online misinformation.
Actions (login required)
 |
View Item |