Learning Deterministic Weighted Automata with Queries and Counterexamples

Part of Advances in Neural Information Processing Systems 32 (NeurIPS 2019)

AuthorFeedback »Bibtex »Bibtex »MetaReview »Metadata »Paper »Reviews »Supplemental »

Authors

Gail Weiss, Yoav Goldberg, Eran Yahav

Abstract

<p>We present an algorithm for reconstruction of a probabilistic deterministic finite automaton (PDFA) from a given black-box language model, such as a recurrent neural network (RNN). The algorithm is a variant of the exact-learning algorithm L*, adapted to work in a probabilistic setting under noise. The key insight of the adaptation is the use of conditional probabilities when making observations on the model, and the introduction of a variation tolerance when comparing observations. When applied to RNNs, our algorithm returns models with better or equal word error rate (WER) and normalised distributed cumulative gain (NDCG) than achieved by n-gram or weighted finite automata (WFA) approximations of the same networks. The PDFAs capture a richer class of languages than n-grams, and are guaranteed to be stochastic and deterministic -- unlike the WFAs.</p>