Semantic disambiguation of taxonomies

Polysemy is one of the most difficult problems when dealing with natural language resources. Consequently, automated ontology learning from textual sources (such as web resources) is hampered by the inherent ambiguity of human language. In order to tackle this problem, this paper presents an automatic and unsupervised method for disambiguating taxonomies (the key component of a final ontology). It takes into consideration the amount of resources available in the Web as the base for inferring information distribution and semantics. It uses co-occurrence analysis and clustering techniques in order to group those taxonomical concepts that belong to the same “sense”. The final results are automatically evaluated against WordNet synsets.