Whisper-Taiwanese model V0.5 (Tv0.5): 這個模型是由國立臺南大學執行國科會產學合作計畫，使用 openai/whisper-large-v3-turbo 微調的版本，並執行國科會TAIDE台英語家庭先導計畫，與真平出版社合作，使用中小學教材內容及學生學習資料進行模型微調，用於真平教材台語辨識。並與國研院國網中心合作，運用國網中心算力以及TAIDE模型，共同建構中小學台語AI學習模型。

Model Details

• Base Model: OpenAI whisper-large-v3-turbo

• Fine-tuned for: 台灣台語語音辨識 (ASR)

• Fine-tuning Framework: Hugging Face Transformers

• Training Duration: 使用兩片 V100，大約 180 小時

• Dataset: 自訂資料集、教育部臺灣台語常用詞辭典，大約 90 小時的資料

• Input Format: 16kHz mono WAV

• License: CC BY-NC 4.0

Step 1: Download the Sample Code from here

Step 2: Log in to your Google account

Step 3: Connect to Colab

Step 4: Click on the File menu, then select Upload notebook to upload the sample code

Step 5: Click on the Runtime menu and select Change runtime type. Set it to Python 3 and T4GPU

• Step 5.1: Download Whisper-Taiwanese Tv0.5 Model (NUTN-KWS/Whisper-Taiwanese-model-v0.5) from Hugging Face

• Step 5.2: Learn how to speak Taiwanese by referring to this webpage 

• Step 5.3: Speak in Taiwanese and record your audio or upload this sample audio to experience

• Step 5.4: Execute ASR (Automatic Speech Recognition) to convert your Taiwanese audio to Chinese text by Tv0.5 Model

• Step 5.5: Download the Chinese-to-English translation model (Helsinki-NLP/opus-mt-zh-en) from Hugging Face

• Step 5.6: Translate the Chinese text into English text by the opus-mt-zh-en Model

Step 6: End