Grammar as a Behavioral Biometric: Using Cognitively Motivated Grammar Models for Authorship Verification

The project is a fresh academic implementation (0 stars, 7 days old) of a paper focusing on Authorship Verification (AV). While its focus on 'cognitively motivated grammar' and explainability addresses a real pain point in digital forensics—where black-box models are often inadmissible in court—the technical moat is shallow. Grammar-based stylometry is a well-established field (e.g., JGAAP, PAN competitions). The defensibility is low because the 'simpler method' described is inherently easy to replicate once the paper's feature set (likely POS-tag n-grams or dependency patterns) is known. The high fork-to-star ratio (6 forks, 0 stars) suggests the developers or a specific research group are active, but there is no broader community adoption. Frontier labs (OpenAI/Anthropic) pose a medium risk; while they don't focus on AV specifically, their LLMs are increasingly capable of both mimicking styles and detecting stylistic anomalies, which could render traditional grammar-based biometrics obsolete within 1-2 years. The project's primary value lies in its specific focus on legal/forensic explainability, a niche larger players currently ignore.