LCMA-SRT: Language-Conditional Mixture-of-Experts Adapters for Joint Multilingual Speech Recognition and Translation

Neural transducers offer an alignment-free framework for speech-to-text modeling, and hierarchical transducer architectures further improve multilingual joint automatic speech recognition (ASR) and speech translation (ST) by stacking a translation-focused encoder on top of an ASR encoder. However, extending hierarchical transducers to multilingual many-to-many settings remains challenging: fully shared models often suffer from negative transfer and unstable target-language generation, while training separate models for each direction is computationally prohibitive. We propose LCMA-SRT (Language-Conditional Mixture-of-Experts Adapters for Speech Recognition and Translation), which augments a hierarchical transducer with language-conditional Mixture-of-Experts (MoE) adapters. A source-conditioned MoE adapter (SRC-MoE) uses source-language embeddings to reduce cross-language interference and improve multilingual ASR. A target-conditioned MoE adapter (TGT-MoE) uses the desired target language to reduce cross-target interference and stabilize target-language generation in many-to-many ST. Experiments on Europarl-ST (9 languages, 72 directions) show that LCMA-SRT improves both ASR and ST within a single joint model, reducing average WER and improving BLEU and COMET over strong hierarchical transducer baselines.

Main Results

Multilingual ASR Pretraining

Model	WER (%) ↓
	de	en	es	fr	it	nl	pl	pt	ro	Avg
CR-CTC	24.57	18.59	20.76	19.24	17.33	36.75	25.28	19.82	18.77	22.35
+ MoE	24.39	18.41	20.16	18.61	17.28	36.83	24.36	19.70	18.79	22.06
+ S-Bias	23.89	17.60	19.58	17.41	16.73	34.72	23.63	18.21	17.97	21.08
+ SRC-MoE	23.34	17.45	19.41	17.34	16.27	35.20	23.28	18.16	17.48	20.88

Many-to-Many Joint Training (Average)

Model	WER (%)↓	Average BLEU ↑
		de	en	es	fr	it	nl	pl	pt	ro	Avg
HENT-SRT-M20×9	23.28	10.7	21.2	19.1	18.2	14.2	16.5	7.2	18.4	12.1	15.3
HENT-SRT-M2M	16.65	2.6	12.8	5.5	4.0	1.8	3.5	1.2	4.9	2.5	4.3
LCMA-SRT	15.71	15.2	25.9	25.8	24.7	20.0	20.5	10.7	23.9	17.6	20.5

Model	LMR (%)↓	Average COMET ↑
		de	en	es	fr	it	nl	pl	pt	ro	Avg
HENT-SRT-M20×9	0.65	0.507	0.656	0.587	0.542	0.565	0.558	0.550	0.609	0.598	0.575
HENT-SRT-M2M	84.95	0.380	0.543	0.478	0.427	0.435	0.401	0.385	0.471	0.406	0.436
LCMA-SRT	0.75	0.574	0.715	0.682	0.627	0.656	0.613	0.616	0.693	0.678	0.651

Usage

1. Install dependencies

# Clone the code repository
git clone https://github.com/linanjie0820/LCMA-SRT.git
cd LCMA-SRT

# Install icefall environment
# See: https://k2-fsa.github.io/icefall/installation/index.html
pip install safetensors sentencepiece torch torchaudio kaldifeat k2 lhotse

2. Download model

pip install huggingface_hub
huggingface-cli download linanjie0820/lcma-srt --local-dir ./model

3. Inference

# ASR (speech recognition)
python model/inference.py --audio test.wav --task asr --src-lang en

# AST (speech translation, e.g., English -> German)
python model/inference.py --audio test.wav --task ast --src-lang en --tgt-lang de

# Both ASR + AST simultaneously
python model/inference.py --audio test.wav --task both --src-lang en --tgt-lang de

Example Output

ASR only:

$ python inference.py --audio en_5525.flac --task asr --src-lang en

==================================================
Decoding ASR...
[ASR] mister president one of the key issues in the new lisbon treaty is the increased role of the european union in the world
==================================================

AST only (English -> German):

$ python inference.py --audio en_5525.flac --task ast --src-lang en --tgt-lang de

==================================================
Decoding AST (en -> de)...
[AST en->de] herr präsident eine wichtige des neuen vertrag von lissabon ist die erhöhten der europäischen union in der welt
==================================================

Both ASR + AST:

$ python inference.py --audio en_5525.flac --task both --src-lang en --tgt-lang de

==================================================
Decoding ASR...
[ASR] mister president one of the key issues in the new lisbon treaty is the increased role of the european union in the world
Decoding AST (en -> de)...
[AST en->de] herr präsident eine wichtige des neuen vertrag von lissabon ist die erhöhten der europäischen union in der welt
==================================================

Loading weights in Python

from safetensors.torch import load_file

# Load weights
state_dict = load_file("model/lcma-srt.safetensors")

# Build model (requires LCMA-SRT code repo)
model.load_state_dict(state_dict, strict=False)
model.eval()

Files

File	Description
lcma-srt.safetensors	Model weights (averaged over training)
config.json	Model architecture configuration
bpe_asr.model	SentencePiece tokenizer for ASR
bpe_ast.model	SentencePiece tokenizer for AST
tokens_asr.txt	ASR token list
tokens_ast.txt	AST token list
inference.py	Inference example script

Training & Decoding

Supported Languages

The dataset covers 9 European languages:

Code	Language
es	Spanish
de	German
en	English
fr	French
nl	Dutch
pl	Polish
pt	Portuguese
ro	Romanian
it	Italian

All non-identical language pairs are processed (72 pairs total).

Output Format

Each final CutSet manifest entry (MonoCut) contains:

{
  "id": "en_5525-758",
  "start": 0,
  "duration": 6.76,
  "channel": 0,
  "supervisions": [
    {
      "id": "en_5525",
      "recording_id": "en_5525",
      "start": 0.0,
      "duration": 6.76,
      "channel": 0,
      "text": "mister president one of the key issues in the new lisbon treaty is the increased role of the european union in the world",
      "language": "en",
      "speaker": "unknown",
      "custom": {
        "st_text": "herr präsident einer der hauptpunkte des neuen vertrages von lissabon ist die aufwertung der rolle der europäischen union in der welt",
        "lang": "de"
      }
    }
  ],
  "recording": {
    "id": "en_5525",
    "sources": [
      {
        "type": "file",
        "channels": [0],
        "source": "data/Europarl-ST/audio/test/en_5525.flac"
      }
    ],
    "sampling_rate": 16000,
    "num_samples": 108160,
    "duration": 6.76
  },
  "type": "MonoCut"
}

Key fields:

• supervisions[0].text: ASR ground truth transcription
• supervisions[0].language: source language
• supervisions[0].custom.st_text: ST ground truth translation
• supervisions[0].custom.lang: target language

For training and full decoding scripts, please refer to the GitHub repository.

Citation

@inproceedings{li2026lcma,
  title={LCMA-SRT: Language-Conditional Mixture-of-Experts Adapters for Joint Multilingual Speech Recognition and Translation},
  author={Li, Nanjie and Guo, Xiaoyong and Huang, Hao and Haihua, Xu and Shi, Wei},
  booktitle={Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  pages={35363--35377},
  year={2026}
}

License

Apache License 2.0