shape

Global tech firm improves recognition of user intent in 96 language variants

Global tech firm improves its virtual assistant’s ability to give users helpful and accurate answers.
shape dots shape dots

The company wanted to train its virtual assistant to generate more accurate and helpful information for end users, improving its performance and language coverage. 

It needed example user queries translated into 96 different language variants, phrased in multiple different ways, all synonymous with the intent of the source. 

TrainAI by RWS developed an efficient, scalable operational model to provide services including query creation, annotation and ranking, quality control of tasks and mature operations management. 

In the five years that the project has been running, TrainAI has delivered over 5 million queries and worked with the company to improve the accuracy of the data labelling.

shape

In the five years that the project has been running, TrainAI has delivered over 5 million queries and worked with the company to improve the accuracy of the data labelling.

Challenges

  • Improve virtual assistant performance in 96 language variants 
  • Generate more helpful and accurate information for users

Results

  • Over 5 million queries delivered in 5 years 
  • 96 languages covered 
  • Tangible process improvements 
  • Accuracy and efficiency gains