Determining Semantic Relatedness through the Measurement of Discrimination Information Using Jensen Difference

Measurement of semantic relatedness has been addressed in a number of application tasks and by researchers in a variety of disciplines. Measurement of discrimination information of terms is a fundamental issue for many areas of science. In this study, we attempt to introduce relatedness measures based on discrimination measures, with the aim of making fundamental concepts accessible and usable to the broad community of data analysis practitioners.We present an in-depth investigation into the basic concept of discrimination information conveyed in a term based on Jensen difference. The discrimination measures can then naturally and conveniently be utilized to introduce two generic concepts of semantic relatedness. We also address the issue of estimating arguments embedded in the relatedness measures and then demonstrate how our method can be supported by empirical evidence drawn from performance experiments.