Chatbot Training Data | OXO Innovation
globe-americas EN FR PT ES

Chatbot Training Data: The Challenge
(PDF Version)

In the spring of 2020, OXO was hired to provide machine learning data for a Silicon Valley tech giant. One of the GAFA companies needed a Quebec French dataset to localize its white-label customer service chatbot.

The project involved a total of 4,500 customer service scenarios for different industries—scenarios like, “You just bought a plane ticket and you want to know if pets are allowed on the airline,” or “You want to change cellphone providers but you’re worried about losing your number.” For every scenario, two people had to produce 55 utterances each using tags and keywords, for a total of 110 unique utterances per scenario.

The client needed to produce a Canadian French dataset to localize its chatbot. However, the Quebec labour market is relatively small and difficult to recruit for, which is why they called on OXO’s local expertise.

To complete the project within a one-month deadline, we had to recruit, test and manage 120 people. People who were available to start right away, who were able to write in idiomatic French, and who had the creativity to come up with 55 ways of asking the same question. They also had to be able to do so quickly (churning out around 55 utterances per hour) to keep the project on schedule and on budget.
 OXO recruited over a hundred people and completed half a million utterances within two weeks.

The Solution

Our Human Relations Manager reached out to our large network of collaborators as well as local universities and encouraged staff to spread the message among their personal contacts by offering a referral bonus. Although machine learning was uncharted territory for us at this point, we realized that we already had the robust vendor management structure in place to pull this recruitment challenge off. We also had the necessary linguistic expertise: our internal team of Quebec French translators and revisers were able to test all applicants and run QA.

OXO recruited over a hundred people and completed half a million utterances within two weeks. We finished the project to the client’s satisfaction within 1 month. This success gave us the confidence to launch a new division, OXO AI, which builds on OXO’s main strengths to provide companies with top-quality, localized machine learning data.

Our AI Experts

oxo-employee-picture Denise Rodrigues Language Lead
 | Rio de Janeiro
oxo-employee-picture Marc-Antoine Quintin Language Specialist
 | Montreal
oxo-employee-picture Stéphanie Levesque Language Specialist
 | Sherbrooke
oxo-employee-picture Eduardo Gama Localization Engineer
 | Rio de Janeiro

Explore More Case Studies

Metro Brands
retail
The supermarket chain Metro partnered with OXO to translate thousands of product descriptions, or "romance copy," for its two private labels. Learn how our translations were tailored to seduce English-speaking consumers.
explore
Lightspeed
software
Lightspeed is a software company with users in roughly 100 countries—creating serious need for software localization. Learn how OXO's translation services help Lightspeed conquer new markets around the globe.
explore