Update README.md
Browse files
README.md
CHANGED
|
@@ -99,6 +99,11 @@ dataset_info:
|
|
| 99 |
|
| 100 |
The master dataset can be found at http://hdl.handle.net/11356/1785.
|
| 101 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 102 |
The ParlaSpeech-CZ dataset is built from the transcripts of parliamentary proceedings available in the Czech part of the ParlaMint corpus (http://hdl.handle.net/11356/1859), and the parliamentary recordings available from the AudioPSP dataset (https://hdl.handle.net/11234/1-5404).
|
| 103 |
|
| 104 |
The dataset consists of audio segments that correspond to specific sentences in the transcripts. The transcript contains word-level alignments to the recordings, each instance consisting of character and millisecond start and end offsets, allowing for simple further segmentation of long sentences into shorter segments for ASR and other memory-sensitive applications. Sequences longer than 30 seconds have already been removed from this dataset, which should allow for a simple usage on most modern GPUs.
|
|
|
|
| 99 |
|
| 100 |
The master dataset can be found at http://hdl.handle.net/11356/1785.
|
| 101 |
|
| 102 |
+
<div style="border: 5px solid #ff6700; padding: 10px; margin: 10px 0;">
|
| 103 |
+
<strong>Notice:</strong> ParlaSpeech corpora are currently in the process of enrichment with new features. Follow our progress here: <a href="http://clarinsi.github.io/parlaspeech">http://clarinsi.github.io/parlaspeech</a>
|
| 104 |
+
</div>
|
| 105 |
+
|
| 106 |
+
|
| 107 |
The ParlaSpeech-CZ dataset is built from the transcripts of parliamentary proceedings available in the Czech part of the ParlaMint corpus (http://hdl.handle.net/11356/1859), and the parliamentary recordings available from the AudioPSP dataset (https://hdl.handle.net/11234/1-5404).
|
| 108 |
|
| 109 |
The dataset consists of audio segments that correspond to specific sentences in the transcripts. The transcript contains word-level alignments to the recordings, each instance consisting of character and millisecond start and end offsets, allowing for simple further segmentation of long sentences into shorter segments for ASR and other memory-sensitive applications. Sequences longer than 30 seconds have already been removed from this dataset, which should allow for a simple usage on most modern GPUs.
|