Scraping the documents with llm to generate numerical metrics and applying statistics on numbers

OP likely means statistical regressions.