Tracing the Evolution of Approaches to Semantic Similarity Analysis

Weronika Adrian, Sebastian Skoczeń, Szymon Majkut, Krzysztof Kluza, Antoni Ligęza


Capturing the essence of semantic similarity of words or concepts in order to quantify it and measure has been an inspiring challenge for the last decades. From corpus-based statistics to metrics based on structured knowledge bases, a plethora of methods has been proposed in several branches of Artificial Intelligence. Recently, with the advent of knowledge graphs, a renewed interest in similarity metrics can be observed. Choosing appropriate metrics that will work best in a given situation is not a trivial task. To help navigate through the semantic similarity algorithms and understand the characteristics of them, we have analyzed the fundamental proposals in this domain and the evolution of them over the years. In this paper, we present a review of the approaches to measuring semantic similarity of entities in knowledge bases. We organize the findings into a taxonomy and analyze the relations between and within the identified categories. To complement the research with a practical solution, we present a new tool that supports the literature review process with graph-based and temporal visualizations.


Paper Citation