No abstract

This paper presents a technique to interpret and visualize intermediate layers in generative CNNs trained on raw speech data in an unsupervised manner. We argue that averaging over feature maps after ReLU activation in each transpose convolutional layer yields interpretable time-series data. This technique allows for acoustic analysis of intermediate layers that parallels the acoustic analysis of human speech data: we can extract F0, intensity, duration, formants, and other acoustic properties from intermediate layers in order to test where and how CNNs encode various types of information. We further combine this technique with linear interpolation of a model's latent space to show a causal relationship between individual variables in the latent space and activations in a model's intermediate convolutional layers. In particular, observing the causal effect between linear interpolation and the resulting changes in intermediate layers can reveal how individual latent variables get transformed into spikes in activation in intermediate layers.We train and probe internal representations of two modelsa bare WaveGAN architecture and a ciwGAN extension which forces the Generator to output informative data and results in the emergence of linguistically meaningful representations. Interpretation and visualization is performed for three basic acoustic properties of speech: periodic vibration (corresponding to vowels), aperiodic noise vibration (corresponding to fricatives), and silence (corresponding to stops). The proposal also allows testing of higher-level morphophonological alternations such as reduplication (copying). In short, using the proposed technique, we can analyze how linguistically meaningful units in speech get encoded in each convolutional layer of a generative neural network.

This paper presents a technique to interpret and visualize intermediate layers in CNNs trained on raw speech data in an unsupervised manner. We show that averaging over feature maps after ReLU activation in each convolutional layer yields interpretable time-series data. The proposed technique enables acoustic analysis of intermediate convolutional layers. To uncover how meaningful representation in speech gets encoded in intermediate layers of CNNs, we manipulate individual latent variables to marginal levels outside of the training range. We train and probe internal representations on two models -a bare WaveGAN architecture and a ciwGAN extension which forces the Generator to output informative data and results in emergence of linguistically meaningful representations. Interpretation and visualization is performed for three basic acoustic properties of speech: periodic vibration (corresponding to vowels), aperiodic noise vibration (corresponding to fricatives), and silence (corresponding to stops). We also argue that the proposed technique allows acoustic analysis of intermediate layers that parallels the acoustic analysis of human speech data: we can extract F0, intensity, duration, formants, and other acoustic properties from intermediate layers in order to test where and how CNNs encode various types of information. The models are trained on two speech processes with different degrees of complexity: a simple presence of [s] and a computationally complex presence of reduplication (copied material). Observing the causal effect between interpolation and the resulting changes in intermediate layers can reveal how individual variables get transformed into spikes in activation in intermediate layers. Using the proposed technique, we can analyze how linguistically meaningful units in speech get encoded in different convolutional layers.

Comparing artificial neural networks with outputs of neuroimaging techniques has recently seen substantial advances in (computer) vision and text-based language models. Here, we propose a framework to compare biological and artificial neural computations of spoken language representations and propose several new challenges to this paradigm. The proposed technique is based on a similar principle that underlies electroencephalography (EEG): averaging of neural (artificial or biological) activity across neurons in the time domain, and allows to compare encoding of any acoustic property in the brain and in intermediate convolutional layers of an artificial neural network. Our approach allows a direct comparison of responses to a phonetic property in the brain and in deep neural networks that requires no linear transformations between the signals. We argue that the brain stem response (cABR) and the response in intermediate convolutional layers to the exact same stimulus are highly similar without applying any transformations, and we quantify this observation. The proposed technique not only reveals similarities, but also allows for analysis of the encoding of actual acoustic properties in the two signals: we compare peak latency (i) in cABR relative to the stimulus in the brain stem and in (ii) intermediate convolutional layers relative to the input/output in deep convolutional networks. We also examine and compare the effect of prior language exposure on the peak latency in cABR and in intermediate convolutional layers. Substantial similarities in peak latency encoding between the human brain and intermediate convolutional networks emerge based on results from eight trained networks (including a replication experiment). The proposed technique can be used to compare encoding between the human brain and intermediate convolutional layers for any acoustic property and for other neuroimaging techniques.

Comparing artificial neural networks (ANNs) with outputs of brain imaging techniques has recently seen substantial advances in (computer) vision and text-based language models. Here, we propose a framework to compare biological and artificial neural computations of spoken language representations and propose several new challenges to this paradigm. Using a technique proposed by Begus and Zhou (2021b), we can analyze encoding of any acoustic property in intermediate convolutional layers of an artificial neural network. This allows us to test similarities in speech encoding between the brain and artificial neural networks in a way that is more interpretable than the majority of existing proposals that focus on correlations and supervised models. We introduce fully unsupervised deep generative models (the Generative Adversarial Network architecture) trained on raw speech to the brain-and-ANN-comparison paradigm, which enable testing of both the production and perception principles in human speech. We present a framework that parallels electrophysiological experiments measuring complex Auditory Brainstem Response (cABR) in human brain with intermediate layers in deep convolutional networks. We compared peak latency in cABR relative to the stimulus in the brain stem experiment, and in intermediate convolutional layers relative to the input/output in deep convolutional networks. We also examined and compared the effect of prior language exposure on the peak latency in cABR, and in intermediate convolutional layers of a phonetic property. Specifically, the phonetic property (i.e., VOT =10 ms) is perceived differently by English vs. Spanish speakers as voiced (e.g. [ba]) vs voiceless (e.g. [pa]). Critically, the cABR peak latency to the VOT phonetic property is different between English and Spanish speakers, and peak latency in intermediate convolutional layers is different between English-trained and Spanish-trained computational models. Substantial similarities in peak latency encoding between the human brain and intermediate convolutional networks emerge based on results from eight trained networks (including a replication experiment). The proposed technique can be used to compare encoding between the human brain and intermediate convolutional layers for any acoustic property.