Study: Back-to-basics approach can match or outperform AI in language analysis
6 hours ago
- #grammar-based-method
- #authorship-analysis
- #transparent-AI
- A grammar-based method named LambdaG can match or exceed AI systems in authorship verification.
- LambdaG uses grammatical patterns, function words, and sentence structure to create a writer's statistical profile.
- It was tested on 12 real-world datasets, including emails and forums, often outperforming neural network models.
- The approach offers greater transparency and lower computational costs compared to many AI systems.
- Applications include forensic linguistics, criminal investigations, and online abuse detection.