Information theoretic measures for clusterings comparison: is a correction for chance necessary?