Abduction Algorithms for Grammar Discovery.
Abstract
A model of languages syntax acquisition is formulated as an inference problem: to guess the wiring diagram of an unreliable automaton. A constructive method is developed which solves the grammatical inference problem via an abductory inductive process applied to the sample strings generated by the stochastic automaton whose internal wiring diagram is unavailable for inspection. The right invariant equivalence classes which correspond to the states of the sought-for automaton are established by the training sequence and a teacher. The structural description of strings is found directly without a priori assumptions on the number of states (or lengths of strings).
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 1977
- Accession Number
- ADA044442
Entities
People
- Stefan Shrier
Organizations
- Brown University