Abduction Machines and Language Acquisition.
Abstract
This paper deals with a model of language syntax acquisition. It is assumed that the artificial language has a finite description which we hope to discover on the basis of a finite samples of sentences. Specifically excluded is the simple formulation of the observations actually made, although a naive description is highly desirable. In the terminology of learning models, an insightful model is to be preferred over the rote learning exemplified in a list. The various guises and disguises of this problem are found in artificial intelligence, human cognitive studies, pattern recognition, linguistics and in systems theory under labels such as inductive inference, automation identification and grammatical inference. As far as modelling of acquisition is concerned, little attention has been focussed on possible solutions. The following investigation is motivated toward the presentation of model models. If we restrict consideration to regular grammars, the number of possible grammars is overwhelmingly large for even small size alphabets and modest numbers of variables. This implies that models for analogy to real world phenomena which exhibit language acquisition ability cannot be based on enumerative inference or finite search techniques.
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 01, 1976
- Accession Number
- ADA037127
Entities
People
- S. Shrier
Organizations
- Brown University