Normalized Relevance Distance – A Stable Metric for Computing Semantic Relatedness over Reference Corpora
Christoph Schaefer, Daniel Hienert, Thomas Gottron · Frontiers in artificial intelligence and applications · 2014
We propose the Normalized Relevance Distance (NRD): a robust metric for computing semantic relatedness between terms. NRD makes use of a controlled reference corpus for a statistical analysis. The analysis is based on the relevance scores and joint occurrence of terms in documents. On the basis of established reference datasets, we demonstrate that NRD does not require sophisticated data tuning and is less dependent on the choice of the reference corpus than comparable approaches.