# The Hidden Cost of Digits: Number Normalization and WER in ASR Systems

arXiv: https://arxiv.org/abs/2609.21084 · HTML: https://mlsanity.com/paper/2609.21084/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 13 of its 22 references that are in the map.
- First posted on arXiv 17 Sep 2026.

## Builds on (most cited first, 13 of 13)

- [Robust Speech Recognition via Large-Scale Weak Supervision](https://mlsanity.com/paper/2212.04356/) (2022)
- [VoxPopuli: A Large-Scale Multilingual Speech Corpus for Representation Learning, Semi-Supervised Learning and Interpretation](https://mlsanity.com/paper/2101.00390/) (2021)
- [Earnings-21: A Practical Benchmark for ASR in the Wild](https://mlsanity.com/paper/2104.11348/) (2021)
- [Canary-1B-v2 & Parakeet-TDT-0.6B-v3: Efficient and High-Performance Models for Multilingual ASR and AST](https://mlsanity.com/paper/2509.14128/) (2025)
- [RNN Approaches to Text Normalization: A Challenge](https://mlsanity.com/paper/1611.00068/) (2016)
- [NeMo Inverse Text Normalization: From Development To Production](https://mlsanity.com/paper/2104.05055/) (2021)
- [OWSM v4: Improving Open Whisper-Style Speech Models via Data Scaling and Cleaning](https://mlsanity.com/paper/2506.00338/) (2025)
- [Comparison of Multilingual Self-Supervised and Weakly-Supervised Speech Pre-Training for Adaptation to Unseen Languages](https://mlsanity.com/paper/2305.12606/) (2023)
- [WER we are and WER we think we are](https://mlsanity.com/paper/2010.03432/) (2020)
- [BESTOW: Efficient and Streamable Speech Language Model with the Best of Two Worlds in GPT and T5](https://mlsanity.com/paper/2406.19954/) (2024)
- [What is lost in Normalization? Exploring Pitfalls in Multilingual ASR Model Evaluations](https://mlsanity.com/paper/2409.02449/) (2024)
- [NEST: Self-supervised Fast Conformer as All-purpose Seasoning to Speech Processing Tasks](https://mlsanity.com/paper/2408.13106/) (2024)
- [Handling Numeric Expressions in Automatic Speech Recognition](https://mlsanity.com/paper/2408.00004/) (2024)

## Abstract (arXiv)

Modern automatic speech recognition (ASR) systems trained on extremely large datasets can produce transcripts with numbers written in Arabic numerals. This creates a need for fair comparison with models that output verbatim texts and proper processing of reference transcripts. Popular approaches often reduce text normalization to lowercase and remove punctuation, with no additional normalization applied to languages other than English. In this work, we analyze the impact of normalization of numerical expressions in the evaluation of ASR systems in various languages, using Polish as an example of a highly inflective language. We perform experiments on VoxPopuli and The Polish Parliamentary speech datasets and estimate word error rate (WER) differences for different text normalization approaches. We show that the difference due to the lack of number normalization in WER may be substantial - more than 2 percentage points, and often higher than the differences between systems in popular multilingual benchmarks.