javaabu/dhivehi-majlis-speech

Dataset Card for Dhivehi Majlis Speech 1.0

Dataset Summary

Dhivehi Majlis Speech is a Dhivehi speech dataset created from data annotated by Javaabu Pvt. Ltd..

The dataset contains around 10.5 hrs of speech collected from parliament sessions at The Peoples Majlis of Maldives (Maldivian Parliament) consisting of audio from different MPs from 6 different sessions.

Supported Tasks and Leaderboards

• Automatic Speech Recognition
• Text-to-Speech

Languages

Dhivehi

Dataset Structure

Data Instances

A typical data point comprises the path to the audio file and its sentence.

{
  'path': 'dhivehi-majlis-speech-train/waves/majlis_jalsa_24_11_2021_p9_73.wav',  
  'sentence': 'މިއަދު އަންނަމުންދާ ތަރައްޤީއަކީ، އަޅުގަނޑުމެންގެ ރައްޔިތުން އެތައް ދުވަހެއް ވަންދެން ކުރަމުންދިޔަ އުންމީދުތައް',
  'audio': {
    'path': 'dhivehi-majlis-speech-train/waves/majlis_jalsa_24_11_2021_p9_73.wav', 
    'array': array([-0.00048828, -0.00018311, -0.00137329, ...,  0.00079346, 0.00091553,  0.00085449], dtype=float32), 
    'sampling_rate': 16000
  }, 
}

Data Fields

• path (string): The path to the audio file.

• sentence (string): The transcription for the audio file.

• audio (dict): A dictionary containing the path to the downloaded audio file, the decoded audio array, and the sampling rate. Note that when accessing the audio column: dataset[0]["audio"] the audio file is automatically decoded and resampled to dataset.features["audio"].sampling_rate. Decoding and resampling of a large number of audio files might take a significant amount of time. Thus it is important to first query the sample index before the "audio" column, i.e. dataset[0]["audio"] should always be preferred over dataset["audio"][0].

Data Splits

The speech material has been subdivided into portions for train, test and validation.

	Train	Validation	Test	Total
Utterances	4647	580	580	5807
Duration	08:23:06	01:20:28	01:03:55	10:47:28

Dataset Creation

Curation Rationale

[More Information Needed]

Source Data

Initial Data Collection and Normalization

Data was provided by The Peoples Majlis of Maldives. For transcript normalization, new lines and multiple whitespace has been replaced by single spaces. Some of the text written in numerals and arabic has been transliterated to Thaana. However, some arabic strings and numerals may still remain.

Who are the source language producers?

[More Information Needed]

Annotations

Annotation process

The audio was annotated using Audino

Who are the annotators?

Annotation was done by Javaabu staff and hired annotators. These annotators were:

• Hassan Ulvan Mohamed
• Zayan Saudhulla
• Ibrahim Shareef
• Mohamed Jailam

Personal and Sensitive Information

[More Information Needed]

Considerations for Using the Data

Social Impact of Dataset

[More Information Needed]

Discussion of Biases

[More Information Needed]

Other Known Limitations

[More Information Needed]

Additional Information

Dataset Curators

[More Information Needed]

Licensing Information

[More Information Needed]

Citation Information

@misc{Javaabu_2023, 
    title = "Dhivehi Majlis Speech Dataset", 
    url = "https://huggingface.co/datasets/javaabu/dhivehi-majlis-speech", 
    journal = "Hugging Face",
    author = {{Javaabu Pvt. Ltd.}}, 
    year = "2023", 
    month = jul
} 

Contributions

• Arushad Ahmed
• Mohamed Jailam
• Ibrahim Shareef