1. Insights
  2. AI Data
  3. Case Study

Virtual assistant training in 14 languages

Discover how TELUS International edits and enhances a leading virtual assistant’s voice grammar framework for one of the world’s largest technology corporations.

case study thumbnail
  • Share on X
  • Share on Facebook
  • Share via email

14
computational linguists

4000
hours of work completed

14
languages/locales

The challenge

Our client is a multinational technology company that develops, manufactures, licenses and sells computer software, consumer electronics and related services. As a part of their product offering, they offer a voice-controlled virtual assistant that can perform a range of tasks, including the ability to schedule appointments and set reminders.

Now a major feature of most smartphones and tablets, virtual assistants are becoming increasingly integral to daily life. Virtual assistants are able to perform a variety of helpful actions based on user input, such as setting calendar events, sending emails and playing music.

As one of the world’s largest technology corporations, our client’s virtual assistant serves over 150 million active users per month. Continuously improving their assistant’s performance is of critical importance to increasing their global market share. To do this, the company needed a partner that could train, test and scale their virtual assistant software in over a dozen new languages.

The TELUS International solution

Our client partnered with TELUS International because of our linguistic and technical expertise, and our ability to provide high-quality language processing services. Key factors of the solution include:

  • Data annotation: Our project managers facilitate pronunciation checks, validate transcription and generate pronunciation tasks for the client’s speech recognition system.
  • Data collection: To test the client’s virtual assistant technology, we generate thousands of sentence variations on a monthly basis. These are later recorded by native speakers before input into the system.
  • Voice grammar creation: A voice grammar extensible markup language (XML) contains the set of rules that define what virtual assistants can recognize as meaningful input. Our team of computational linguists works within the client’s framework to transpose complex grammar rules into 14 locales.

The results

During this engagement, our team of 14 computational linguists completed over 4,000 hours of work. By partnering with TELUS International, our client was able to train its virtual assistant to better understand and respond to user queries in 14 languages.

Following the project’s initial success, the client expanded our involvement within the greater company, and has repeatedly commended the team for its strong performance and consistent quality.


Check out our solutions

Power your NLP algorithms using our accurately annotated AI training data.

Learn more