SemSimp is a parametric method for evaluating the semantic similarity of digital resources that is based on the notion of information content. It exploits a weighted reference ontology of concepts and requires resources to be semantically annotated, each by means of a set of concepts from the ontology. Specifically, the weights of the concepts can be calculated either by considering the available annotations or only the structure of the ontology. SemSimp was evaluated against six representative semantic similarity methods proposed in the literature. Experiments were run on a large real-world dataset based on the Association for Computing Machinery (ACM) digital library, including both a statistical analysis and an expert judgment assessment. The main result shows that the SemSimp annotation frequency configuration, when combined with the geometric average normalization factor, outperforms the other methods.
SemSimp: A Parametric Method for Evaluating the Semantic Similarity of Digital Resources
De Nicola A.;
2024-01-01
Abstract
SemSimp is a parametric method for evaluating the semantic similarity of digital resources that is based on the notion of information content. It exploits a weighted reference ontology of concepts and requires resources to be semantically annotated, each by means of a set of concepts from the ontology. Specifically, the weights of the concepts can be calculated either by considering the available annotations or only the structure of the ontology. SemSimp was evaluated against six representative semantic similarity methods proposed in the literature. Experiments were run on a large real-world dataset based on the Association for Computing Machinery (ACM) digital library, including both a statistical analysis and an expert judgment assessment. The main result shows that the SemSimp annotation frequency configuration, when combined with the geometric average normalization factor, outperforms the other methods.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
