INDUCTIVE INFERENCE FROM NOISY EXAMPLES USING THE HYBRID FINITE-STATEFILTER

Recurrent neural networks processing symbolic strings can be regarded as adaptive neural parsers. Given a set of positive and negative examp les, picked up from a given language, adaptive neural parsers can effe ctively be trained to infer the language grammar. In this paper we use adaptive neural parsers to face the problem of inferring grammars fro m examples that are corrupted by a kind of noise that simply changes t heir membership, We propose a training algorithm, referred to as hybri d finite state filter (HFF), which is based on a parsimony principle t hat penalizes the development of complex rules. We report very promisi ng experimental results showing that the proposed inductive inference scheme is indeed capable of capturing rules, while removing noise.