# Normalization through Fine-tuning: Understanding Wav2vec 2.0 Embeddings for Phonetic Analysis

arXiv: https://arxiv.org/abs/2503.04814 · HTML: https://mlsanity.com/paper/2503.04814/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 8 of its 20 references that are in the map.
- First posted on arXiv 4 Mar 2025.

## Builds on (most cited first, 8 of 8)

- [wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations](https://mlsanity.com/paper/2006.11477/) (2020)
- [AISHELL-1: An Open-Source Mandarin Speech Corpus and A Speech Recognition Baseline](https://mlsanity.com/paper/1709.05522/) (2017)
- [Insights on representational similarity in neural networks with canonical correlation](https://mlsanity.com/paper/1806.05759/) (2018)
- [Unsupervised Speech Recognition](https://mlsanity.com/paper/2105.11084/) (2021)
- [What all do audio transformer models hear? Probing Acoustic Representations for Language Delivery and its Structure](https://mlsanity.com/paper/2101.00387/) (2021)
- [Probing Acoustic Representations for Phonetic Properties](https://mlsanity.com/paper/2010.13007/) (2020)
- [Neural Representations for Modeling Variation in Speech](https://mlsanity.com/paper/2011.12649/) (2020)
- [Automatic recognition of suprasegmentals in speech](https://mlsanity.com/paper/2108.01122/) (2021)

## Built on by (most cited first, 1 of 1)

- [Probing for Phonology in Self-Supervised Speech Representations: A Case Study on Accent Perception](https://mlsanity.com/paper/2506.17542/) (2025)

## Abstract (arXiv)

Phonetic normalization plays a crucial role in speech recognition and analysis, ensuring the comparability of features derived from raw audio data. However, in the current paradigm of fine-tuning pre-trained large transformer models, phonetic normalization is not deemed a necessary step; instead, it is implicitly executed within the models. This study investigates the normalization process within transformer models, especially wav2vec 2.0. Through a comprehensive analysis of embeddings from models fine-tuned for various tasks, our results demonstrate that fine-tuning wav2vec 2.0 effectively achieves phonetic normalization by selectively suppressing task-irrelevant information. We found that models fine-tuned for multiple tasks retain information for both tasks without compromising performance, and that suppressing task-irrelevant information is not necessary for effective classification. These findings provide new insights into how phonetic normalization can be flexibly achieved in speech models and how it is realized in human speech perception.