French Speech Recordings From Mobile Phone Guiding Scenarios

Aggregating 10.9 hours of French speech from 401 speakers, with each speaker contributing 50 sentences. The audio was recorded using mainstream Android phones and iPhones for guiding scenarios including in-car, smart home, and smart speech assistant contexts. Texts were manually transcribed for accuracy.

Use Cases

• Train an automatic speech recognition (ASR) model on manually transcribed French speech from 401 speakers.
• Fine-tune a speech model for in-car or smart home command scenarios using the described guiding context data.
• Analyze speaker variation or accent patterns across the 401-speaker corpus of mobile-recorded French.

Strengths

• Dataset contains 10.9 hours of audio, providing a substantial amount of speech data.
• Recordings involve 401 distinct speakers, offering potential for speaker diversity.
• Texts are manually transcribed, ensuring high label accuracy for speech recognition tasks.
• Uses mainstream mobile phones (Android and iPhone), reflecting real-world recording conditions.

Limitations

• The dataset is described as a sample of a larger paid dataset, so the available scope may be limited.
• Specific audio quality metrics, speaker demographics, or geographic origin of speakers are not provided.
• No information on file formats, sampling rates, or audio channels is available for technical assessment.

Provenance

Source

Nexdata

Collection Method

Recorded by 401 speakers using mainstream Android phones and iPhones; texts manually transcribed.

Freshness

Last updated on 2025-04-23.