Contextual lexical methods outperform raw LLM prompting for Hindi pregroup assignment (64.56% accuracy), while symbolic lexical repair improves LLM outputs. This advances scalable grammatical type annotation for multilingual quantum NLP pipelines.
@informaq.bsky.social