Dataset Viewer
Auto-converted to Parquet Duplicate
Search is not available for this dataset
audio
audio

French Call Center Speech Dataset - 860 Hours

Description

A large-scale dataset of real French call center conversations.

The dataset contains 5,500 audio recordings with a total duration of approximately 860.19 hours.

It covers multiple customer-service and business domains and can be used for automatic speech recognition (ASR), speech-to-text, conversational AI, dialogue systems, call center analytics, and speech research.

Dataset Summary

  • Language: French
  • Number of audio files: 5,500
  • Total duration: ~860.19 hours
  • Format: WAV
  • Codec: PCM 16-bit (pcm_s16le)
  • Channels: Mono (1 channel)
  • Sample rate: 8,000 Hz
  • Bit depth: 16-bit
  • Bitrate: ~128 kbps

Domain Distribution

Domain Files Hours Share
Energy / Utilities 1,718 341.45 39.69%
Telecommunications 1,125 156.21 18.16%
General / Other Conversations 752 106.68 12.40%
Customer Service / Support 584 92.20 10.72%
Sales / Telemarketing 569 92.10 10.71%
Home Renovation / Housing 378 36.22 4.21%
Government Aid / Grants 294 28.45 3.31%
Non-conversational / Unclear 63 5.52 0.64%
Contracts / Billing 9 0.83 0.10%
Logistics 5 0.30 0.03%
Legal / Advisory 1 0.12 0.01%
Healthcare 1 0.06 <0.01%
Financial Services 1 0.05 <0.01%

Metadata

Metadata is available for the recordings and includes:

  • audio_file
  • duration_sec
  • channels
  • format
  • codec
  • bit_rate
  • bit_depth
  • sample_rate_hz
  • Domain

Potential Use Cases

  • Automatic Speech Recognition (ASR)
  • Speech-to-text
  • Call center speech analytics
  • Conversational AI
  • Dialogue systems
  • Domain classification
  • Customer-service analytics

License

Commercial license only.

Redistribution, resale, sublicensing, or public hosting of the full dataset is not permitted.

Contact

Email: marat@menkroft.com
Telegram: @Marat_DV


Description en français

Ce jeu de données contient des conversations réelles de centres d'appels en français.

Il comprend 5 500 enregistrements audio pour une durée totale d'environ 860,19 heures.

Le corpus couvre plusieurs secteurs d'activité et scénarios de service client. Il peut être utilisé pour la reconnaissance automatique de la parole (ASR), le speech-to-text, l'IA conversationnelle, les systèmes de dialogue et l'analyse des centres d'appels.

Résumé du jeu de données

  • Langue : Français
  • Nombre de fichiers audio : 5 500
  • Durée totale : ~860,19 heures
  • Format : WAV
  • Codec : PCM 16-bit (pcm_s16le)
  • Canaux : Mono (1 canal)
  • Fréquence d'échantillonnage : 8 000 Hz
  • Résolution : 16-bit
  • Débit : ~128 kbps

Principaux domaines

  • Énergie et services publics
  • Télécommunications
  • Conversations générales / autres
  • Service client / assistance
  • Ventes / télémarketing
  • Rénovation / logement
  • Aides publiques / subventions
  • Contrats / facturation
  • Logistique
  • Autres domaines

Utilisations possibles

  • Reconnaissance automatique de la parole (ASR)
  • Speech-to-text
  • Analyse des conversations de centres d'appels
  • IA conversationnelle
  • Systèmes de dialogue
  • Classification par domaine
  • Analyse du service client

Licence

Utilisation commerciale uniquement.

La redistribution, la revente, la sous-licence ou l'hébergement public du jeu de données complet sont interdits.

Contact

Email: marat@menkroft.com
Telegram: @Marat_DV


Sample

Downloads last month
63