Fake news detection: deep semantic representation with enhanced feature engineering.

Area of Science:

• Natural Language Processing
• Machine Learning
• Computational Linguistics

Background:

• Social media facilitates the rapid spread of fake news, impacting public and governmental spheres.
• Existing misinformation detection models primarily rely on semantic features from deep contextualized text representations.
• There is a need for more robust methods to accurately identify fake news.

Discussion:

• This research introduces a novel deep neural architecture integrating content-based feature engineering with semantic features.
• The model utilizes parallel convolutional neural network (CNN) layers for semantic feature extraction, followed by a fully connected layer for classification.
• The effectiveness of combining diverse features is investigated for improved fake news detection.

Key Insights:

• The proposed model significantly improves accuracy and F1-score on both English COVID-19 and Persian fake news datasets compared to baseline models.
• Performance gains were observed against state-of-the-art methods in both datasets.
• Content-based features demonstrably enhance the performance of semantic models in fake news detection.

Outlook:

• Future research could explore additional content-based features for further performance gains.
• The model's applicability to other misinformation domains and languages warrants investigation.
• Further development could lead to more effective tools for combating online misinformation.