Datasets:
Update README.md
Browse files
README.md
CHANGED
|
@@ -23,8 +23,7 @@ evaluation, and comparison of ASR systems, speech representation models, and mul
|
|
| 23 |
|
| 24 |
KasaSpeech is suitable for:
|
| 25 |
|
| 26 |
-
* Automatic Speech Recognition (ASR)
|
| 27 |
-
* Code-switching speech recognition
|
| 28 |
* Text-To-Speech (TTS)
|
| 29 |
* Multilingual speech modeling
|
| 30 |
* Speech representation learning
|
|
@@ -88,8 +87,6 @@ Example transcript:
|
|
| 88 |
Me phone no a-crack-i, henfa na mɛtumi a-fix-i screen no?
|
| 89 |
```
|
| 90 |
|
| 91 |
-
---
|
| 92 |
-
|
| 93 |
## Dataset Creation
|
| 94 |
|
| 95 |
### Collection Process
|
|
@@ -104,8 +101,6 @@ All recordings were manually transcribed following standardized annotation guide
|
|
| 104 |
|
| 105 |
The dataset includes recordings from speakers spanning multiple age groups and genders. Speaker identities have been anonymized using unique identifiers.
|
| 106 |
|
| 107 |
-
---
|
| 108 |
-
|
| 109 |
## Dataset Characteristics
|
| 110 |
|
| 111 |
* **Total recordings:** 54,855
|
|
@@ -115,21 +110,6 @@ The dataset includes recordings from speakers spanning multiple age groups and g
|
|
| 115 |
* **Recording style:** Prompted, natural code-switched speech
|
| 116 |
* **Transcriptions:** Human-annotated
|
| 117 |
|
| 118 |
-
---
|
| 119 |
-
|
| 120 |
-
## Intended Uses
|
| 121 |
-
|
| 122 |
-
KasaSpeech is intended to support research in:
|
| 123 |
-
|
| 124 |
-
* Automatic Speech Recognition (ASR)
|
| 125 |
-
* Code-switching speech modeling
|
| 126 |
-
* Low-resource multilingual speech processing
|
| 127 |
-
* African language technologies
|
| 128 |
-
* Benchmarking multilingual speech models
|
| 129 |
-
* Fine-tuning and evaluating speech foundation models such as Whisper, wav2vec 2.0, MMS, and SeamlessM4T
|
| 130 |
-
|
| 131 |
-
---
|
| 132 |
-
|
| 133 |
## Limitations
|
| 134 |
|
| 135 |
* Demographic representation may not be perfectly balanced across speaker groups.
|
|
@@ -137,7 +117,6 @@ KasaSpeech is intended to support research in:
|
|
| 137 |
* The dataset primarily reflects Ghanaian English–Twi code-switching and may not generalize to all Akan dialects or other multilingual contexts.
|
| 138 |
* Although carefully curated, minor transcription inconsistencies may remain.
|
| 139 |
|
| 140 |
-
---
|
| 141 |
|
| 142 |
## Citation
|
| 143 |
|
|
@@ -152,14 +131,6 @@ If you use **KasaSpeech** in your work, please cite:
|
|
| 152 |
}
|
| 153 |
```
|
| 154 |
|
| 155 |
-
---
|
| 156 |
-
|
| 157 |
-
## Licensing
|
| 158 |
-
|
| 159 |
-
Please refer to the dataset repository for the applicable license.
|
| 160 |
-
|
| 161 |
-
---
|
| 162 |
-
|
| 163 |
## Contact
|
| 164 |
|
| 165 |
For questions, bug reports, or collaboration opportunities, please open a discussion on the Hugging Face dataset page. Contributions, feedback, and research collaborations are welcome.
|
|
|
|
| 23 |
|
| 24 |
KasaSpeech is suitable for:
|
| 25 |
|
| 26 |
+
* Automatic Speech Recognition (ASR) for Code-switching speech
|
|
|
|
| 27 |
* Text-To-Speech (TTS)
|
| 28 |
* Multilingual speech modeling
|
| 29 |
* Speech representation learning
|
|
|
|
| 87 |
Me phone no a-crack-i, henfa na mɛtumi a-fix-i screen no?
|
| 88 |
```
|
| 89 |
|
|
|
|
|
|
|
| 90 |
## Dataset Creation
|
| 91 |
|
| 92 |
### Collection Process
|
|
|
|
| 101 |
|
| 102 |
The dataset includes recordings from speakers spanning multiple age groups and genders. Speaker identities have been anonymized using unique identifiers.
|
| 103 |
|
|
|
|
|
|
|
| 104 |
## Dataset Characteristics
|
| 105 |
|
| 106 |
* **Total recordings:** 54,855
|
|
|
|
| 110 |
* **Recording style:** Prompted, natural code-switched speech
|
| 111 |
* **Transcriptions:** Human-annotated
|
| 112 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 113 |
## Limitations
|
| 114 |
|
| 115 |
* Demographic representation may not be perfectly balanced across speaker groups.
|
|
|
|
| 117 |
* The dataset primarily reflects Ghanaian English–Twi code-switching and may not generalize to all Akan dialects or other multilingual contexts.
|
| 118 |
* Although carefully curated, minor transcription inconsistencies may remain.
|
| 119 |
|
|
|
|
| 120 |
|
| 121 |
## Citation
|
| 122 |
|
|
|
|
| 131 |
}
|
| 132 |
```
|
| 133 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 134 |
## Contact
|
| 135 |
|
| 136 |
For questions, bug reports, or collaboration opportunities, please open a discussion on the Hugging Face dataset page. Contributions, feedback, and research collaborations are welcome.
|