Ontological foundations for experimental science knowledge bases
Natalya Fridman Noy, Carole D. Hafner · Applied Artificial Intelligence · 2000
The framework for representing domain ontologies presented in this paper extends existing ontological models and traditional frame-based formalisms. This work was motivated by the representational challenges posed by the domains of experimental sciences (biology, chemistry, physics) and the task of intelligent text retrieval. A detailed ontology for the field of experimental molecular biology is presented, which is used to illustrate the need for and application of the features of the framework. An extended frame-based formalism is defined to support these features. The ability of the framework to support intelligent retrieval from a knowledge base of molecular-biology research papers is demonstrated by providing answers to queries that could not be fully answered using previous approaches. The extensions to ontological framework include : category conversions, processes that change the category or identity of their participants; object histories to track substances through a series of experimental processes, including category conversions; object complexes, temporary configurations of objects with properties of their own; and process complexes, groups or sequences of interrelated actions that comprise an experimental technique or procedure. Features of the frame-based formalism include: slot groups for identifying sets of relations that license common inferences; and open-filler classes that combine knowledge of likely slot values with the ability to handle unexpected values. Evaluation techniques that are used to assess the adequacy of the ontology are presented: the ontology's conceptual coverage of the domain, its potential usefulness in improving the quality of query answering, and its formal consistency and reusability by the knowledge-sharing community are evaluated.