Inducing Regular Grammars Using Recurrent Neural Networks
Computation and Language
2018-06-27 v2
Abstract
Grammar induction is the task of learning a grammar from a set of examples. Recently, neural networks have been shown to be powerful learning machines that can identify patterns in streams of data. In this work we investigate their effectiveness in inducing a regular grammar from data, without any assumptions about the grammar. We train a recurrent neural network to distinguish between strings that are in or outside a regular language, and utilize an algorithm for extracting the learned finite-state automaton. We apply this method to several regular languages and find unexpected results regarding the connections between the network's states that may be regarded as evidence for generalization.
Cite
@article{arxiv.1710.10453,
title = {Inducing Regular Grammars Using Recurrent Neural Networks},
author = {Mor Cohen and Avi Caciularu and Idan Rejwan and Jonathan Berant},
journal= {arXiv preprint arXiv:1710.10453},
year = {2018}
}
Comments
Accepted to L&R 2018 workshop, ICML & IJCAI