Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,57 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language:
|
| 3 |
+
- id
|
| 4 |
+
base_model:
|
| 5 |
+
- openai/whisper-base
|
| 6 |
+
pipeline_tag: automatic-speech-recognition
|
| 7 |
+
datasets:
|
| 8 |
+
- mozilla-foundation/common_voice_23_0
|
| 9 |
+
---
|
| 10 |
+
|
| 11 |
+
# Whisper Base Model – Indonesian ASR
|
| 12 |
+
|
| 13 |
+
## Model Description
|
| 14 |
+
This model is a fine-tuned version of **openai/whisper-base** for **Automatic Speech Recognition (ASR)** in **Indonesian (id)**.
|
| 15 |
+
It supports transcription of Indonesian speech into text across various audio conditions, with performance and resource usage depending on the selected model size.
|
| 16 |
+
|
| 17 |
+
## Intended Use
|
| 18 |
+
- Indonesian speech-to-text transcription
|
| 19 |
+
- Research and experimentation
|
| 20 |
+
- Educational and academic purposes
|
| 21 |
+
- Application development and benchmarking
|
| 22 |
+
|
| 23 |
+
Model variants (tiny, base, small, medium, large) differ in accuracy, speed, and hardware requirements. Users should select the size that best matches their constraints and objectives.
|
| 24 |
+
|
| 25 |
+
## Limitations
|
| 26 |
+
- Transcription quality depends on audio clarity, speaker accent, and background noise
|
| 27 |
+
- Smaller variants may produce higher error rates on long or complex audio
|
| 28 |
+
- Larger variants require significantly more compute and memory
|
| 29 |
+
- Outputs should be reviewed before use in critical or high-risk applications
|
| 30 |
+
|
| 31 |
+
## Training Data
|
| 32 |
+
This model was fine-tuned using **Mozilla Common Voice v23.0 (Indonesian)**.
|
| 33 |
+
Common Voice is a publicly available, community-driven speech dataset released by Mozilla under a permissive license.
|
| 34 |
+
Dataset characteristics such as speaker diversity, recording quality, and utterance length may influence model behavior.
|
| 35 |
+
|
| 36 |
+
## Evaluation
|
| 37 |
+
The model is typically evaluated using **Word Error Rate (WER)**.
|
| 38 |
+
Evaluation results may vary depending on dataset, domain, audio conditions, and model size.
|
| 39 |
+
|
| 40 |
+
## Training results
|
| 41 |
+
| Step | Training Loss |
|
| 42 |
+
|------|---------------|
|
| 43 |
+
| 100 | 0.880500 |
|
| 44 |
+
| 200 | 0.472300 |
|
| 45 |
+
| 300 | 0.408100 |
|
| 46 |
+
| 400 | 0.328500 |
|
| 47 |
+
| 500 | 0.226000 |
|
| 48 |
+
| 600 | 0.237500 |
|
| 49 |
+
| 700 | 0.148600 |
|
| 50 |
+
| 800 | 0.111600 |
|
| 51 |
+
| 900 | 0.104900 |
|
| 52 |
+
| 1000 | 0.073900 |
|
| 53 |
+
| 1100 | 0.063100 |
|
| 54 |
+
| 1200 | 0.050300 |
|
| 55 |
+
| 1400 | 0.039800 |
|
| 56 |
+
| 1500 | 0.031000 |
|
| 57 |
+
| 1550 | 0.031400 |
|