WIP This repo contains features extraction methods that can be used on text to improve explainability of LLM performances.