doc: segmenting, cleansing desc
Browse files
README.md
CHANGED
|
@@ -9,7 +9,7 @@ tags:
|
|
| 9 |
license: cc0-1.0
|
| 10 |
---
|
| 11 |
|
| 12 |
-
# Farsi Youtube ASR Dataset
|
| 13 |
|
| 14 |
This dataset consists of over **385** hours of audio extracted from various YouTube videos in the Persian language.
|
| 15 |
|
|
@@ -21,9 +21,11 @@ The dataset includes Farsi content from various types of videos spanning from ol
|
|
| 21 |
- Educational Content
|
| 22 |
- Interviews
|
| 23 |
- Documentaries
|
|
|
|
| 24 |
|
| 25 |
## Note
|
| 26 |
-
This dataset contains raw, unvalidated transcriptions.
|
|
|
|
| 27 |
- Perform their own quality assessment
|
| 28 |
- Create their own train/validation/test splits based on their specific needs
|
| 29 |
- Validate a subset of the data if needed for their use case
|
|
|
|
| 9 |
license: cc0-1.0
|
| 10 |
---
|
| 11 |
|
| 12 |
+
# Farsi Youtube 2024 ASR Dataset
|
| 13 |
|
| 14 |
This dataset consists of over **385** hours of audio extracted from various YouTube videos in the Persian language.
|
| 15 |
|
|
|
|
| 21 |
- Educational Content
|
| 22 |
- Interviews
|
| 23 |
- Documentaries
|
| 24 |
+
Utterances and sentences are extracted based on the timing of subtitles.
|
| 25 |
|
| 26 |
## Note
|
| 27 |
+
This dataset contains raw, unvalidated auto-generated transcriptions. Transcriptions may include inaccuracies due to mal-transcriptions, and timing may occasionally be imprecise. Many efforts have been made to cleanse the data using various methods and software.
|
| 28 |
+
Users are advised to:
|
| 29 |
- Perform their own quality assessment
|
| 30 |
- Create their own train/validation/test splits based on their specific needs
|
| 31 |
- Validate a subset of the data if needed for their use case
|