Prof. Mustafa Jarrar, PhD
Professor of Artificial Intelligence
Computer Science, Birzeit University, Palestine
Title: Linguistic Ontologies and Wordnets
Wordnets play an important role in understanding and retrieving unstructured information, especially in NLP and IR tasks. Their importance is also increasing to support managing and retrieving of structured data in new areas, such as Knowledge Graphs, multilingual Big Data, and medical informatics. Such new needs are demanding wordnets to be formal and play the role of ontologies.
The difference between wordnets and ontologies might not be obvious, especially because both have similar structures, e.g. considering synsets as concepts and hyponyms as subsumptions. However, synsets in wordnets are linguistically motivated concepts (i.e. units of thoughts), while concepts in ontologies are classes of instances. Additionally, subsumption is a subset relation, in the extensional or intensional sense, rather than a linguistic general-specific relationship. Furthermore, ontologies are typically application-specific rich axiomatizations, while wordnets are general-purpose mental lexicons, thus axiomatizing them would be a rigidification.
This talk will discuss the notion of linguistic ontology, which can play the role of being a wordnet and an ontology at the same time. The talk will also discuss what can be learned from the ontology engineering literature to build wordnets with ontologically and formally cleaner content.
The second part of the talk will present the Arabic Ontology, which is an Arabic wordnet built with formal and ontological analysis in mind. The ontology is represented in a similar structure as wordnets, and is fully mapped to the Princeton Wordnet, as well as with the WikiData knowledge graph and with many Arabic-multilingual lexicons. The ontology is being built at Birzeit University, in Palestine, and it is available at https://ontology.birzeit.edu/concept/293198