Understanding natural language demands the use of many types of knowledge. For example, one must know definitions of words, rules of syntax, and knowledge about the world.
Some of the world knowledge has an emphasis on intention, i.e., the purpose behind actions. It is this knowledge which allows us to recognize the distinction between requests and orders, promises and threats, or lies and informings.
In order to recognize intention in behavior, whether the behavior be speech or otherwise, it is necessary to have a model of the beliefs of others. Furthermore one needs to know how actions typically fit together into larger units and how different aspects of an action are determined by intentions and beliefs.
This research examines some of the “belief systems knowledge” used in language understanding and generation. It includes a theory of personal causation.
The theory supplies the tools to account for purposeful behavior. Using primitives of the theory the social aspect of an action can be described. The social aspect is that which depends on beliefs and intentions. Patterns of behavior, called
social action paradigms (SAP’s), are then defined in terms cf social actions. The SAP’s provide a structure for episodes analogous to the structure a grammar provides for sentences.
Bruce, Bertram C. (1983). Belief systems and language understanding. In W. A. Sedelow & S. V. Sedelow (Eds.), Computers in language research two: Two parts in one volume (Trends in linguistics. Studies and monographs, 19) (pp. 113-160). New York: de Gruyter. [ISBN13: 9789027930095] Also as BBN Report No. 2973 (1975, January). Cambridge, MA: Bolt Beranek and Newman.
Bruce, Bertram C. (1975). Generation as a social action. In Bonnie L. Nash-Webber & R. C. Schank (Eds.), Theoretical issues in natural language processing (pp. 74-77). Cambridge, MA: Association for Computational Linguistics. Reprinted (1986) in B. J. Grosz, K. S. Jones, and Bonnie L. Webber (Eds.), Natural language processing (pp. 419-422). Los Altos, CA: Morgan Kaufmann.
Bruce, Bertram C., & Schmidt, C. F. (1974, July ). Episode understanding and belief guided parsing (Report No. CBM-TR-32). New Brunswick, NJ: Rutgers University, Computer Science Department.