Speech data collection, grammar development, localization and tuning in 16 languages
A leading automotive manufacturer needed global language resources to tune and enhance performance of in-vehicle speech technology in 16 languages. The manufacturer turned to Appen and our language expertise on six continents to meet their needs.
Appen worked closely with the client team to gain a deep understanding of their requirements and to provide them with the services and resources they needed, including:
- A full set of voice commands and Text-to-Speech prompts for each language, localized from US English.
- Finite State Grammars (needed for accurate Speech Recognition) – multiple FSGs were developed and optimized for each language to accommodate various OEM infotainment systems.
- Recording and transcription of 88 speakers per language, each reading over 500 commands – speaker selection ensured coverage of a broad range of dialects.
- Pronunciation lexicons for each language.
- Dynamic script content across various categories (e.g. phone numbers, album names, addresses etc.) for all languages.
- System performance assessment and refinement.
Appen’s recording hardware and software were customized to satisfy client requirements for microphones and acoustic conditions. Work was done following an agreed-upon schedule enabling the manufacturer to do advanced planning for which language features would make it into specific car models.
This client appreciated Appen’s ability to act as a single point of contact for all language services. As they knew the in-car functionalities would expand over time, both in terms of language coverage and feature breadth, the client extended our collaboration. Together we defined a service ‘library’ from which the client can order services in any language. This reduced overhead for both companies, and enabled the fastest possible turn-around on new projects.