Choosing the Right Word
The information resources comprising the worldwide Internet is a powerful research tool. However, difficulties exist in locating the exact information that one requires due to the abundance of information. There is a well-known trade-off between the quantity and quality of search results. Either one gets many results not so related to their target or fewer results are returned that are more strongly correlated with the search term(s). Ontologies concern text-based descriptions of concepts that can be exploited by search engines to return meaningful results. Researchers at the Austrian Research Institute for Artificial Intelligence (OeFAI) evaluated a number of term utility functions used in deriving ontologies. The results of the comparison single out a term utility function that allows the user to easily specify the relation between recall (i.e. quantity) and precision (i.e. quality). The term utility function performed extremely well when tested with real data and was validated by experts in the field. The application of this result involves practically any and all disciplines, though it is particularly useful for those with extensive, highly evolved terminology. Examples include the fields of biology and medicine. The tool can be used to develop descriptions of single documents (e.g. scientific publications) or collections of documents. It will also be of value to those creating new ontologies or revising existing ontologies. OeFAI is pursuing the promotion of this concept while at the same time working to resolve some instances of unsatisfactory performance.
