Fischer Black, Harvard 1960
It is concluded that the data-base question-answerer has passed from initial research into the early developmental phase. The most difficult and important research questions for the advancement of general-purpose language processors are seen to be concerned with measuring meaning, dealing with ambiguities, translating into formal languages and searching large tree structures. DOI
in [ACM] CACM 8(01) Jan 1965 view details
in Minsky, Marvin (ed.), "Semantic Information Processing" Cambridge, MA: MIT Press 1968 view details
Generally, natural language question-answering systems use some formal internal representation for facts and questions in order to facilitate deductive manipulations. In a number of earlier systems, the representation was based upon some type of limited relational calculi, i.e., Raphaels SIR and Blacks SQA (Raphael, 1968; Black, 1968).
Green and Raphael (1968) subsequently developed a system that offered the full expressiveness of the first-order predicate calculus for the representation of natural language information. The deductive procedure of this system was based on an automatic theorem-proving algorithm that was first described by Robinson (1965) and improved upon by Wos, Robinson, Carson, and Shalla (1967) and others. [See Green, Cordell & Raphael, 1968; Wos, Robinson & Carson, 1964a; Wos, Robinson & Carson, 1964b; Wos, Robinson & Carson, 1965; Wos, Robinson, Carson & Shalla, 1967.]
The use of first-order predicate calculus as a formal language for the representation of natural language information when used in conjunction with automatic theorem-proving procedures was a significant improvement over previous schemes. However, the first-order predicate calculus cannot be used to express relationships between relations or allow variables to range over relations as well as objects.
For example, suppose it is necessary to put into the first-order language the sentences:
John crossed the street after the light changed. (1)
A car must always yield to a pedestrian. (2)
In (1) we are unable to put the sentence into the first-order language because we have a relation, namely after, whose arguments are forced to be relations, namely crossed and changed, rather than individuals. In (2), we cannot put the sentence into the first-order language because we are faced with quantification of a variable that ranges over situations, not individuals. That is, the sentence states that for all possible situations, a certain condition holds (i.e., that a car must yield to a pedestrian).
in Proceedings of the Academy of Information and Management Sciences, 7(2) Las Vegas, 2003 view details