Oromo ASR Data Collection
输入“/”快速插入
Oromo ASR Data Collection
飞书用户55014月15日创建
Intro to Afaan Oromoo (Oromo Language)
•
Native to Ethiopia and Kenya
•
Language with the largest number of native speakers in Ethiopia.
•
4th most spoken African Language in Africa
•
Spoken by around 45 million people in East Africa
~ Source: Wikipedia
Data Collection App
This week I tested the app and fixed some bugs.
◦
Forbidding user to submit audio while recording or playing.
◦
Prevent changing sentence or updating user contribution when audio uplaod is failed.
Added:
Displaying statistics to the admin:
Total contribution. Number of Female and Male registered.
•
Model
User info:
Id, name username, gender, email, ageGroup, Dialect, langrank, accept_terms
List of Recorded Sentences' ID
Sentence:
A dictionary containing:
ID and sentence
Data
Sentences are stored as JSON file within the app as an asset.
Sentence length:
Min: 4 words
Max: 20 words
In total there are 27,996 sentences.
A random sentence is displayed for the user to read. A list of sentence IDs recorded by the user is stored. The sentences with those IDs in the list will not be shown to the user again.
[
{
"id":1,
"sentence":"Kun waan namni hundi beekuudha."
},
{
"id":2,
"sentence":"Harka fuudhaa qorraa dhaan na dubbise."
},
....
]
Records:
The user records are stored on cloud storage with a json file containing the recorded sentence and speaker info and the corresponding audio file in .wav format. The audio file and JSON file have the same name.