Evaluation of an Unsupervised Ontology Enrichment Framework

The paper describes an unsupervised framework for domain ontology enrichment based on mining domain text corpora. Specifically, we enrich the hierarchical backbone of an existing ontology, i.e. its taxonomy, with new domain-specific concepts. The framework is based on hierarchical self-organizing maps. Terms extracted by mining a text corpus encode contextual content information, in a distributional vector space. The enrichment proceeds by populating the existing taxonomy with the extracted terms and attaching them as hyponymsfor the intermediate and leaf nodes of the taxonomy. We propose an evaluation setting in which we asses the power of attraction of the population of terms towards the branches of the taxonomy (recall) and the precision of attaching correct hyponyms (accuracy). The evaluation results presented are in the "four universities" domain.