RelTopic: A Graph-Based Semantic Relatedness Measure in Topic Ontologies and Its Applicability for Topic Labeling of Old Press Articles

Tracking #: 2536-3750

This paper is currently under review
Mirna El Ghosh
Nicolas Delestre
Jean-Philippe Kotowicz
Cecilia Zanni-Merk
Habib Abdulrab

Responsible editor: 
Special Issue Cultural Heritage 2021

Submission type: 
Full Paper
Graph-based semantic measures have been used to solve problems in several domains. They tend to compare ontological entities in order to estimate their semantic similarity or relatedness. While semantic similarity is applicable to hierarchies, semantic relatedness is adapted to ontologies. However, designing semantic relatedness measures is a difficult and challenging issue. In this paper, we propose a novel semantic measure within topic ontologies, named RelTopic, for assessing the relatedness of instances and topics. To design RelTopic, we considered topic ontologies as weighted graphs where topics and instances are represented as weighted nodes and semantic relations as weighted edges. The use of RelTopic is evaluated for labeling old press articles. For this purpose, a topic ontology, named Topic-OPA, is derived from open knowledge graphs by the application of a SPARQL-based fully automatic approach. The ontology building process is based mainly on a set of disambiguated named entities representing the articles. To demonstrate the performance of our approach, a use-case is presented in the context of the old french newspaper Le Matin. Our experiments show that RelTopic produces more than 80% relevant labeling topics as compared to the topics assigned by human annotators.
Full PDF Version: 
Under Review