Unconventional usage of entropy in the field of web usage data preprocessing and machine translation evaluation

Abstract

This paper focuses on an unconventional usage of entropy. On one side it deals with preprocessing phase, especially the session identification using the Reference Length method. Entropy, in this case, offers an alternative to determining the ratio of auxiliary pages that is important for this method. With the approach introduced in this paper, the need of a sitemap becomes void. On the other hand, the paper looks at entropy in the case of reliability analysis of Machine Translation metrics. In this case, entropy offers also an alternative mean to validate the metrics.