VincentGranville/Large-Language-Models

Crawls structured web directories (Wolfram MathWorld) to build hierarchical category structures and word embeddings for enriching LLM prompts across multiple knowledge sources (Wikipedia, ArXiv, Google Scholar, Stack Exchange). Generates domain-specific prompt augmentation by extracting semantic relationships from directory taxonomies and content metadata. Python-based pipeline with preprocessing utilities designed for domain focus areas like Probability & Statistics, supporting downstream integration with GPT and search APIs.

472 stars. No commits in the last 6 months.