Beyond Words: A Mathematical Framework for Interpreting Large Language Models