A global technology company was in need of speech data to support their speech research and development team.
The work involved several tiers including 100 hour-long in-person monologues and almost 100 multi-person conversational speech sessions amongst two, three, and four participants at a time. These conversations then required, transcription, and intricate utterance “start and stop” tagging.
The “Total Package” in Speech Collection
iSoftStone implemented a two-phase solution for completing this project. The conversational speech collection took place locally through recruiting participants from several sources. The recordings used a variety of different microphones, headphones, and a customized recording application. Once recordings were completed, the conversational files were transferred to our global crowds, who reviewed recording quality and tagged all audio files with start and stop times for each individual speaker in every conversation.
On Time, On Budget, and Quality that’s On Point
All conversational audio files, transcriptions, and tagged start and stop files were delivered to the client ahead of the agreed upon delivery date. The delivered files met the clients SLA’s for quality at 98%. The client was able to use this data to enhance their data models that create the “human like personal assistance” commonly seen in products and devices across the technology market today.