A Self-Attention Mechanism-Based Model for Early Detection of Fake News
Bibliographic record
Abstract
Extensive studies have indicated that fake news has become one of the major threats to our social system (e.g., influencing public opinion, financial markets, journalism, and health system), and its impact cannot be understated, particularly in our current socially and digitally connected society. In the past years, this problem has been investigated from different perspectives and various disciplines, such as computer science, political science, information science, and linguistics. Even though such efforts have proposed many helpful solutions, it remains challenging to detect fake news in its early phases of dissemination. Based on previously reported studies, detecting fake news early after its propagation is a very tough task due to the unavailability of context-based features within the first hours of spreading and the ineffectiveness of merely content-based features methods. To address this challenge, we propose a new framework for detecting fake news in the early stages of its propagation. The first three components of the proposed framework convert each news article’s propagation network into a sequence of nodes after preprocessing and feature extraction. The last module of our framework leverages a self-attention mechanism-based encoder. Self-attention technique is the core of the well-known transformer model, which has achieved promising results in different areas, especially in complex tasks such as language translation. In the module, a new representation of the input sequence is generated, which is mapped to a label for the news article by a binary classifier. We evaluated our method on two datasets and achieved promising results. The achieved F1 scores by the proposed model on the GossipCop and PolitiFact datasets are higher than the best baseline model by 9% and 6%, respectively.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.004 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.001 | 0.002 |
| Open science | 0.003 | 0.001 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".