Extracting finite structure from infinite language
This paper presents a novel unsupervised neural network model for learning the finite-state properties of an input language from a set of positive examples. The model is demonstrated to learn the Reber grammar perfectly from a randomly generated training set and to generalize to sequences beyond the length of those found in the training set. Crucially, it does not require negative examples. 30% of the tests yielded perfect grammar recognizers, compared with only 2% reported by other authors for simple recurrent networks. The paper was initially presented at AI-2004 conference where it won the Best Technical Paper award.
Citation : McQueen, T. et al. (2005) Extracting finite structure from infinite language. Knowledge-Based Systems, 18(4-5), pp. 135-141.
ISSN : 0950-7051
Research Group : Centre for Computational Intelligence