Update README.md
Browse files
README.md
CHANGED
|
@@ -33,7 +33,23 @@ The training set comprised 650,000 sentence pairs per target language (English,
|
|
| 33 |
|
| 34 |
## Evaluation
|
| 35 |
|
| 36 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 37 |
|
| 38 |
## Usage
|
| 39 |
|
|
|
|
| 33 |
|
| 34 |
## Evaluation
|
| 35 |
|
| 36 |
+
The reported values are Phone Error Rate (PER) across the [Wikipron](https://github.com/CUNY-CL/wikipron) dataset.
|
| 37 |
+
|
| 38 |
+
Compared Models/Frameworks:
|
| 39 |
+
|
| 40 |
+
- [eSpeak NG](https://github.com/espeak-ng/espeak-ng)
|
| 41 |
+
- [Gruut](https://pypi.org/project/gruut)
|
| 42 |
+
- [fdemelo/g2p-mbyt5-12l-ipa-childes-espeak](https://huggingface.co/fdemelo/g2p-mbyt5-12l-ipa-childes-espeak)
|
| 43 |
+
- [OLaPh Framework] (https://github.com/iisys-hof/olaph)
|
| 44 |
+
- OLaPh LLM (this)
|
| 45 |
+
|
| 46 |
+
| Language | espeak | gruut | byt5 | olaph | olaph_llm |
|
| 47 |
+
| :--- | :--- | :--- | :--- | :--- | :--- |
|
| 48 |
+
| **de** | 0.17594 | 0.17558 | 0.35496 | 0.04302 | 0.13518 |
|
| 49 |
+
| **en_uk** | 0.14117 | 0.19174 | 0.52258 | 0.08749 | 0.16321 |
|
| 50 |
+
| **en_us** | 0.14588 | 0.16545 | 0.42466 | 0.10491 | 0.14713 |
|
| 51 |
+
| **es** | 0.04324 | 0.04210 | 0.31214 | 0.02582 | 0.03179 |
|
| 52 |
+
| **fr** | 0.06203 | 0.04045 | 0.11421 | 0.03143 | 0.08591 |
|
| 53 |
|
| 54 |
## Usage
|
| 55 |
|