Analyzing analytical methods: The case of phonology in neural models of spoken language
This work addresses the need for systematic evaluation of analysis techniques in NLP and speech processing, though it is incremental as it focuses on a specific case of phonology.
The study systematically compared diagnostic classifiers and representational similarity analysis to assess how neural models encode phonemes and phoneme sequences, finding that using randomly-initialized models is crucial and global-scope methods yield more consistent results.
Given the fast development of analysis techniques for NLP and speech processing systems, few systematic studies have been conducted to compare the strengths and weaknesses of each method. As a step in this direction we study the case of representations of phonology in neural network models of spoken language. We use two commonly applied analytical techniques, diagnostic classifiers and representational similarity analysis, to quantify to what extent neural activation patterns encode phonemes and phoneme sequences. We manipulate two factors that can affect the outcome of analysis. First, we investigate the role of learning by comparing neural activations extracted from trained versus randomly-initialized models. Second, we examine the temporal scope of the activations by probing both local activations corresponding to a few milliseconds of the speech signal, and global activations pooled over the whole utterance. We conclude that reporting analysis results with randomly initialized models is crucial, and that global-scope methods tend to yield more consistent results and we recommend their use as a complement to local-scope diagnostic methods.