This research investigates deception, especially lies of omission, using Natural Language Processing (NLP) techniques. A dataset of 876,784 samples was curated by merging a large-scale fake news dataset and scraped news headlines from the Twitter handle of ‘Times of India’. The study presents a novel multi-task learning pipeline for deception detection, achieving a high F1 score of 0.87. The research also explores the relationship between lies of omission and propaganda techniques, revealing significant correlations. The researchers will make the SEPSIS models and dataset available under the MIT License for further research.
Publication date: 1 Dec 2023
Project Page: https://arxiv.org/abs/2312.00292v1
Paper: https://arxiv.org/pdf/2312.00292
