Can We Still Hear the Accent? Investigating the Resilience of Native Language Signals in the LLM Era

This project is an academic research artifact associated with an ArXiv paper. With 0 stars and 2 forks, it currently serves primarily as a reproducibility package for the study rather than a tool intended for production use. The defensibility is low (2) because the core value lies in the research findings and the labeled dataset rather than a novel, protected technical moat; any NLP researcher could replicate the fine-tuning process on the ACL Anthology. Frontier risk is low because Labs like OpenAI or Anthropic are focused on general-purpose intelligence and alignment, not the sociological study of author L1 backgrounds in specific academic niches. The primary competition comes from other academic groups studying 'stylistic homogenization' or 'AI-generated text detection.' While the 'accent' detection in text is an interesting niche of stylometry, the utility is largely forensic or academic, meaning it faces little risk of platform domination but also has a limited commercial ceiling. The 1-2 year displacement horizon reflects the rapid evolution of LLM capabilities; as models get better at mimicking native-level prosody and syntax, the 'fingerprints' this tool seeks to detect may genuinely disappear, rendering the current classifier obsolete.