Audio Data Collection for Training a Speech Recognition System

Introduction

We were tasked with supporting a project aimed at developing a voice recognition and assistance system. The goal was to collect voice data from native French speakers to train a voice assistant capable of interacting in a natural, fluid, and intuitive manner. The client required a large volume of recordings meeting strict quality standards. We therefore provided comprehensive support to structure and oversee each phase of the data collection process.

Solution and Benefits

• Recording of 1,000 sentences per participant via the client’s mobile app
• Use of predefined scripts supplied and uploaded directly by the client
• Quality checks to ensure recordings were made in quiet environments
• Standardization of collected data to enable effective training of the voice assistant

Results and Conclusions

  • Project completed in just over 3 months
  • Collection of 63,000 voice scripts and over 130,000 hours of recordings
  • All of the client’s qualitative and quantitative requirements fully met

Latest case studies.

    Opps, No posts were found.

Audio Data Collection for Training a Speech Recognition System

Introduction

We were tasked with supporting a project aimed at developing a voice recognition and assistance system. The goal was to collect voice data from native French speakers to train a voice assistant capable of interacting in a natural, fluid, and intuitive manner. The client required a large volume of recordings meeting strict quality standards. We therefore provided comprehensive support to structure and oversee each phase of the data collection process.

Solution and Benefits

• Recording of 1,000 sentences per participant via the client’s mobile app
• Use of predefined scripts supplied and uploaded directly by the client
• Quality checks to ensure recordings were made in quiet environments
• Standardization of collected data to enable effective training of the voice assistant

Results and Conclusions

  • Project completed in just over 3 months
  • Collection of 63,000 voice scripts and over 130,000 hours of recordings
  • All of the client’s qualitative and quantitative requirements fully met

Latest posts.