Automatic Speech Recognition
leduckhai commited on
Commit
587d4cd
·
verified ·
1 Parent(s): b60c723

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +0 -6
README.md CHANGED
@@ -20,14 +20,12 @@ license: mit
20
 
21
  Please refer to newer version which integrates ASR + MT models: [https://huggingface.co/leduckhai/MultiMed-ST](https://huggingface.co/leduckhai/MultiMed-ST)
22
 
23
-
24
  **<div align="center">ACL 2025</div>**
25
 
26
  <div align="center"><b>Khai Le-Duc</b>, Phuc Phan, Tan-Hanh Pham, Bach Phan Tat,</div>
27
 
28
  <div align="center">Minh-Huong Ngo, Chris Ngo, Thanh Nguyen-Tang, Truong-Son Hy</div>
29
 
30
-
31
  > Please press ⭐ button and/or cite papers if you feel helpful.
32
 
33
  <p align="center">
@@ -36,7 +34,6 @@ Please refer to newer version which integrates ASR + MT models: [https://hugging
36
 
37
  * **Abstract:**
38
  Multilingual automatic speech recognition (ASR) in the medical domain serves as a foundational task for various downstream applications such as speech translation, spoken language understanding, and voice-activated assistants. This technology improves patient care by enabling efficient communication across language barriers, alleviating specialized workforce shortages, and facilitating improved diagnosis and treatment, particularly during pandemics. In this work, we introduce MultiMed, the first multilingual medical ASR dataset, along with the first collection of small-to-large end-to-end medical ASR models, spanning five languages: Vietnamese, English, German, French, and Mandarin Chinese. To our best knowledge, MultiMed stands as **the world’s largest medical ASR dataset across all major benchmarks**: total duration, number of recording conditions, number of accents, and number of speaking roles. Furthermore, we present the first multilinguality study for medical ASR, which includes reproducible empirical baselines, a monolinguality-multilinguality analysis, Attention Encoder Decoder (AED) vs Hybrid comparative study and a linguistic analysis. We present practical ASR end-to-end training schemes optimized for a fixed number of trainable parameters that are common in industry settings. All code, data, and models are available online: [https://github.com/leduckhai/MultiMed/tree/master/MultiMed](https://github.com/leduckhai/MultiMed/tree/master/MultiMed).
39
-
40
  * **Citation:**
41
  Please cite this paper: [https://arxiv.org/abs/2409.14074](https://arxiv.org/abs/2409.14074)
42
 
@@ -49,9 +46,6 @@ Please cite this paper: [https://arxiv.org/abs/2409.14074](https://arxiv.org/abs
49
  }
50
  ```
51
 
52
- This repository contains scripts for medical automatic speech recognition (ASR) for 5 languages: Vietnamese, English, German, French, and Mandarin Chinese.
53
- The provided scripts cover model preparation, training, inference, and evaluation processes, based on the dataset *MultiMed*.
54
-
55
  ## Dataset and Pre-trained Models:
56
 
57
  Dataset: [🤗 HuggingFace dataset](https://huggingface.co/datasets/leduckhai/MultiMed), [Paperswithcodes dataset](https://paperswithcode.com/dataset/multimed)
 
20
 
21
  Please refer to newer version which integrates ASR + MT models: [https://huggingface.co/leduckhai/MultiMed-ST](https://huggingface.co/leduckhai/MultiMed-ST)
22
 
 
23
  **<div align="center">ACL 2025</div>**
24
 
25
  <div align="center"><b>Khai Le-Duc</b>, Phuc Phan, Tan-Hanh Pham, Bach Phan Tat,</div>
26
 
27
  <div align="center">Minh-Huong Ngo, Chris Ngo, Thanh Nguyen-Tang, Truong-Son Hy</div>
28
 
 
29
  > Please press ⭐ button and/or cite papers if you feel helpful.
30
 
31
  <p align="center">
 
34
 
35
  * **Abstract:**
36
  Multilingual automatic speech recognition (ASR) in the medical domain serves as a foundational task for various downstream applications such as speech translation, spoken language understanding, and voice-activated assistants. This technology improves patient care by enabling efficient communication across language barriers, alleviating specialized workforce shortages, and facilitating improved diagnosis and treatment, particularly during pandemics. In this work, we introduce MultiMed, the first multilingual medical ASR dataset, along with the first collection of small-to-large end-to-end medical ASR models, spanning five languages: Vietnamese, English, German, French, and Mandarin Chinese. To our best knowledge, MultiMed stands as **the world’s largest medical ASR dataset across all major benchmarks**: total duration, number of recording conditions, number of accents, and number of speaking roles. Furthermore, we present the first multilinguality study for medical ASR, which includes reproducible empirical baselines, a monolinguality-multilinguality analysis, Attention Encoder Decoder (AED) vs Hybrid comparative study and a linguistic analysis. We present practical ASR end-to-end training schemes optimized for a fixed number of trainable parameters that are common in industry settings. All code, data, and models are available online: [https://github.com/leduckhai/MultiMed/tree/master/MultiMed](https://github.com/leduckhai/MultiMed/tree/master/MultiMed).
 
37
  * **Citation:**
38
  Please cite this paper: [https://arxiv.org/abs/2409.14074](https://arxiv.org/abs/2409.14074)
39
 
 
46
  }
47
  ```
48
 
 
 
 
49
  ## Dataset and Pre-trained Models:
50
 
51
  Dataset: [🤗 HuggingFace dataset](https://huggingface.co/datasets/leduckhai/MultiMed), [Paperswithcodes dataset](https://paperswithcode.com/dataset/multimed)