Unlocking Transparent Alignment Through Enhanced Inverse Constitutional AI for Principle Extraction