Temporal reasoning is one of the major issues in several AI fields such as planning, natural language processing, and knowledge representation. In this thesis, a temporal-knowledge representation scheme based on time point is proposed and its reasoning mechanism is developed. Time point is more computationally tractable than time interval.
In the proposed representation scheme, only the primitive facts are maintained in knowledge base and other unspecified temporal relations are deduced from the knowledge base. Basic operations used in the deduction are based on the additions and multiplications of the time point algebra.
Compared with others, the proposed model saves the space usage for maintaining temporal knowledge and reduces the time for updating them when a new temporal relation is asserted. This model, however, requires some extra time for deducing a temporal relation since there are only primitive facts in the knowledge base while others, which use the constraint propagation algorithm, maintain all temporal relations in the database. Accordingly, the proposed model is more useful in the domains, which deal with a number of time points and require frequent acquisitions of new temporal relations.