Solving the problem of polysemy
Polysemy means for a single word or phrase to have multiple, related meanings. In its most systematic forms, this linguistic phenomenon reflects deep conceptual principles according to which we organize our knowledge of the world. Yet, although humans easily process such conceptual relations, they raise notorious difficulties for systems managing linguistically encoded information. The EU-funded ODASP (Ontology driven analysis of nominal systematic polysemy in WordNet) project planned to improve the way information about the world is standardly represented and structured in computer science, viz. in computational ontologies. The project’s primary focus was on ontologies built out of linguistic categories and relationships. The consortium aimed to address the difficulties of representing the conceptual relations underlying systematic polysemy in a major online linguistic database of English called WordNet. In particular, researchers considered the connected problem of multiple-inheritance, which is one of the main sources of the "IS A overloading problem". Project methodologies drew upon philosophical ontology and linguistics. The team identified the proportion of WordNet entries affected by this problem. Researchers further proposed solutions involving a novel kind of semantic relation in WordNet: the ‘constitution relation’. This solution involves enriching WordNet’s structure to properly represent complex concepts, as forming a separate linguistic hierarchy from standard simple concepts. Work yielded a formal theoretical representation of the above relationship within a high-level ontology. Its mapping onto WordNet’s restructured top-level allows for novel semantic inferences. The proposed ontology avoids the limitations of previous proposals by adopting a top-down approach, based on careful analysis of ontological constraints and relations underlying systematic polysemy. Research revealed the conceptual categorisation problem underlying the lexical phenomenon of ‘inherent nominal polysemy’. In so doing, the work showed the relevance of the project’s results for ontologies in general. The project facilitated knowledge-transfer. At the university level, it involved the training of one MSc student in relevant methods. Transfer also occurred at European and international levels, in the form of research collaborations. Beyond the linguistic phenomenon of polysemy, ODASP's categorisation principles may find application in other ontological systems, improving their performance, as argued for the case of biomedical databases.
Keywords
Polysemy, analysis, WordNet, constitution, relation
