The database approach to knowledge representation

Jeffrey David Ullman · 1996

The database theory community, centered around the PODS (Principles of Database Systems) conference has had a long-term interest in logic as a way to represent "data," "information," and "knowledge" (take your pick on the term --- it boils down to facts or atoms and rules, usually Horn clauses). The approach of this community has been "slow and steady," preferring to build up carefully from simple special cases to more general ideas, always paying attention to how efficiently we can process queries and perform other operations on the facts and rules. A powerful theory has developed, and it is beginning to have some impact on applications, especially information-integration engines. Datalog The term Datalog has been coined to refer to Prolog-like rules without function symbols, treated as a logic program. Unlike Prolog, however, the conventional leastfixed -point semantics of the rules is used whenever possible. Example 1: The rules for ancestors can be written as the following Data...

Read the paper · More papers on PaperTik