# Latent Softmax for Data-Efficient Phoneme-Based Multilingual ASR Across Tonal and Non-Tonal Languages

arXiv: https://arxiv.org/abs/2608.01281 · HTML: https://mlsanity.com/paper/2608.01281/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 16 of its 25 references that are in the map.
- First posted on arXiv 2 Aug 2026.

## Builds on (most cited first, 16 of 16)

- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [AISHELL-1: An Open-Source Mandarin Speech Corpus and A Speech Recognition Baseline](https://mlsanity.com/paper/1709.05522/) (2017)
- [Progressive Identification of True Labels for Partial-Label Learning](https://mlsanity.com/paper/2002.08053/) (2020)
- [Universal Phone Recognition with a Multilingual Allophone System](https://mlsanity.com/paper/2002.11800/) (2020)
- [The ASRU 2019 Mandarin-English Code-Switching Speech Recognition Challenge: Open Datasets, Tracks, Methods and Results](https://mlsanity.com/paper/2007.05916/) (2020)
- [Label Relation Graphs Enhanced Hierarchical Residual Network for Hierarchical Multi-Granularity Classification](https://mlsanity.com/paper/2201.03194/) (2022)
- [Whistle: Data-Efficient Multilingual and Crosslingual Speech Recognition via Weakly Phonetic Supervision](https://mlsanity.com/paper/2406.02166/) (2024)
- [Weakly Supervised Representation Learning with Coarse Labels](https://mlsanity.com/paper/2005.09681/) (2020)
- [Multilingual and crosslingual speech recognition using phonological-vector based phone embeddings](https://mlsanity.com/paper/2107.05038/) (2021)
- [CS-Dialogue: A 104-Hour Dataset of Spontaneous Mandarin-English Code-Switching Dialogues for Speech Recognition](https://mlsanity.com/paper/2502.18913/) (2025)
- [Autosegmental Neural Nets: Should Phones and Tones be Synchronous or Asynchronous?](https://mlsanity.com/paper/2007.14351/) (2020)
- [A Review of Learning with Deep Generative Models from Perspective of Graphical Modeling](https://mlsanity.com/paper/1808.01630/) (2018)
- [LLM-based phoneme-to-grapheme for phoneme-based speech recognition](https://mlsanity.com/paper/2506.04711/) (2025)
- [Pronunciation-Lexicon Free Training for Phoneme-based Crosslingual ASR via Joint Stochastic Approximation](https://mlsanity.com/paper/2507.06249/) (2025)
- [Phonemes vs. Projectors: An Investigation of Speech-Language Interfaces for LLM-based ASR](https://mlsanity.com/paper/2604.09332/) (2026)

## Abstract (arXiv)

Phoneme-based multilingual automatic speech recognition (ASR) can share acoustic evidence across languages more directly than language-specific subword modeling. When tonal and non-tonal languages are jointly trained, however, their supervision granularity does not match: tonal languages annotate tone-marked vowels, whereas non-tonal languages typically provide only base-vowel labels. A standard softmax either treats the two as unrelated classes, weakening cross-lingual sharing, or collapses tones, losing distinctions required by tonal languages. We propose Latent Softmax, a connectionist temporal classification (CTC)-compatible output layer that models tone-marked vowels as subclasses and base vowels as major classes, while consonants and the CTC blank remain singleton labels. When only a base-vowel major-class label is observed, the tone-marked vowel subclass is treated as latent and marginalized out. Multilingual experiments on AISHELL-1 Mandarin and LibriSpeech English show that Latent Softmax reduces speech-to-phoneme (S2P) phoneme error rates over a standard softmax multilingual baseline by 8.4% on AISHELL-1, 17.5% on LibriSpeech test-clean, and 12.6% on test-other. The improved speech-to-phoneme encoders also yield consistent word error rate gains for both large-language-model phoneme-to-grapheme conversion and projector-based interfaces. After code-switching adaptation, Latent Softmax further reduces projector-based mixed error rate by 2.6% on ASRU2019 and 9.5% on CS-Dialogue datasets.