# A Technique for Isolating Lexically-Independent Phonetic Dependencies in Generative CNNs

arXiv: https://arxiv.org/abs/2506.09218 · HTML: https://mlsanity.com/paper/2506.09218/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 11 of its 26 references that are in the map.
- First posted on arXiv 10 Jun 2025.

## Builds on (most cited first, 11 of 11)

- [Adversarial Audio Synthesis](https://mlsanity.com/paper/1802.04208/) (2018)
- [SpeechT5: Unified-Modal Encoder-Decoder Pre-Training for Spoken Language Processing](https://mlsanity.com/paper/2110.07205/) (2021)
- [Quantifying Translation-Invariance in Convolutional Neural Networks](https://mlsanity.com/paper/1801.01450/) (2017)
- [CiwGAN and fiwGAN: Encoding information in acoustic data to model lexical learning with Generative Adversarial Networks](https://mlsanity.com/paper/2006.02951/) (2020)
- [Identity-Based Patterns in Deep Convolutional Networks: Generative Adversarial Phonology and Reduplication](https://mlsanity.com/paper/2009.06110/) (2020)
- [How Familiar Does That Sound? Cross-Lingual Representational Similarity Analysis of Acoustic Word Embeddings](https://mlsanity.com/paper/2109.10179/) (2021)
- [Local and non-local dependency learning and emergence of rule-like representations in speech data by Deep Convolutional Generative Adversarial Networks](https://mlsanity.com/paper/2009.12711/) (2020)
- [Basic syntax from speech: Spontaneous concatenation in unsupervised deep neural networks](https://mlsanity.com/paper/2305.01626/) (2023)
- [Exploring How Generative Adversarial Networks Learn Phonological Representations](https://mlsanity.com/paper/2305.12501/) (2023)
- [Disentanglement in a GAN for Unconditional Speech Synthesis](https://mlsanity.com/paper/2307.01673/) (2023)
- [Exploring the encoding of linguistic representations in the Fully-Connected Layer of generative CNNs for Speech](https://mlsanity.com/paper/2501.07726/) (2025)

## Abstract (arXiv)

The ability of deep neural networks (DNNs) to represent phonotactic generalizations derived from lexical learning remains an open question. This study (1) investigates the lexically-invariant generalization capacity of generative convolutional neural networks (CNNs) trained on raw audio waveforms of lexical items and (2) explores the consequences of shrinking the fully-connected layer (FC) bottleneck from 1024 channels to 8 before training. Ultimately, a novel technique for probing a model's lexically-independent generalizations is proposed that works only under the narrow FC bottleneck: generating audio outputs by bypassing the FC and inputting randomized feature maps into the convolutional block. These outputs are equally biased by a phonotactic restriction in training as are outputs generated with the FC. This result shows that the convolutional layers can dynamically generalize phonetic dependencies beyond lexically-constrained configurations learned by the FC.