36. THE CAR WAS FIXED BY MARY.
we will have the following memory configuration:
|37.||CONCEPT =||CN1 = (REPAIR ACTOR nil OBJECT nil)|
|BUFFER =||(CN2 = (CAR)|
|CN3 = (PERSON FIRSTNAME (MARY)))|
During the previous steps, concept CN2 did not get into the ACTOR slot because of the semantic violations and the OBJECT slot because of the syntactic violations. Similarly, CN3 did not fill the ACTOR slot because of syntax and the OBJECT slot because of semantics. However, from the properties of the concept REPAIR, we know that the filler of the ACTOR slot must be a human and the filler of the OBJECT slot must be a physical object. On the basis of these properties alone, we can put CN2 and CN3 into the right slots. This is done at the last step. Of course, there is a price to pay for ignorance. Without knowledge of the passive, the sentence JOHN WAS HIT BY MARY may be understood incorrectly.
The flexibility of the framework described in this chapter is due to the combination of predictive mechanisms with expert modules. Expectations are the main driving force of this type of conceptual analyzer. They control the appli cation of contextual knowledge accessed during the analysis. Expert modules embody the analyzer's general knowledge about major natural-language constructions. They take over control when the predictive mechanisms fail to account for a piece of input. This framework enables us to combine limited knowledge of the language with our model of the text topic to extract a reasonable amount of information from natural-language texts.
Birnbaum L. A., & Selfridge M. G. Problems in conceptual analysis of Platural language (Computer Science Research Report No. 168). Yale University, New Haven, October 1979.
Carbonell J. G. Subjective understanding: Computer models of belief systems (Computer Science Research Report No. 150). Yale University, New Haven, January 1979.
DeJong G. F. Skimming stories in real time: An experiment in integrated understanding (Computer Science Research Report No. 150). Yale University, New Haven, May 1979.
Gershman A. V. "Conceptual analysis of noun groups in english". Proceedings of the 5th International Joint Conference on Artificial Intelligence. Cambridge, Mass., August 1977.
Gershman A. . Knowledge-based parsing (Computer Science Research Report No. 156). Yale University, New Haven, April 1979.
Gershman A. V. Automatic yellow pages assistant-a preliminary report (TM 80-1274-14). Bell Laboratories, October 1980.
Ginsparg J. M. Natural language processing in an automatic programming domain. Doctoral Dissertation (Computer Science Department Report STAN-CS-78-671). Stanford University, Stanford, Calif., June 1978.