Text-to-Ontology Mapping via Natural Language Processing with Application to Search for Relevant Ontologies in Catalysis

The paper presents a machine-learning based approach to text-to-ontology mapping. We explore possibility of matching texts the relevant ontologies using combination artificial neural networks and classifiers. Ontologies are formal specifications shared conceptualizations application domains. While describing same domain, different might be created by domain experts. To enhance reasoning data handling concepts in scientific papers, finding best fitting ontology regarding description contained text corpus. presented this work attempts solve selection representative paragraph from set which used as set. Then, pre-trained fine-tuned Transformer, is embedded into vector space. Finally, becomes classified with respect its relevance selected target ontology. construct embeddings, we experiment training pipelines for natural language processing models. Those embeddings turn later task result assessed compressing visualizing latent space exploring mappings between fragments database chosen ontologies. confirm differences behavior proposed mapper models, test five statistical hypotheses about their relative performance on classification. categorize output classifiers considered. These are, detail, Support Vector Machine (SVM), k-Nearest Neighbor, Gaussian Process, Random Forest, Multilayer Perceptron. Application these concerning catalysis research respective ontologies, suitability evaluated, where was achieved SVM classifier.