Developing first “fairy tale” skill for Google Assistant using Natural Language Processing and Neural Networks.

Vostok AI team was approached by large FMCG brand with product line of baby food. In order to increase customer loyalty and help younger parents to take care of their newborn they proposed the idea to develop a “fairy tale skill” for voice assistant, who will tell popular children’s audio stories. The voice assistant can read fairy tales and call the child by name to keep him/her engaged. The skills become a hit during the Christmas season with 10,000 downloads per month.

 

Challenge.

Our customer tried few other technologies (including chatbot), but none of them worked as children couldn’t focus long on a text info, thus they decided to try voice assistants including Google Assistant. Another challenge was to integrate child’s name into the “fairy tale” context to keep him or her engaged through the story.

Solution.

In order to make this skill's experience unique for each child, we have integrated almost 1,000 of common names into it. Once name is called ,then application will ask child some questions along the way. If the name is not recognized (happens in 3% of cases) then system will call child a “baby”

When child select the book, then reading will start. Every moment child can stop the book and ask to read another one.

asian-little-child-singing-with-happiness-home-with-headphones-microphone.jpg

Project Results.

The project was quite well received by target group (young parents in their 30s) and soon this skill was published on official Google channel. Within first 3 months skill generated 20,000 users (average 6,000 per month) and reached up to 10,000 users in Christmas season. This skill was also called “Fastest Growing Skill” in Russian Google Assistant.

In future we plan to reposition this application not as “bedtime stories” skill, but rather “breakfast assistant”, so we can integrate more interactive scenarios and story lines.

Previous
Previous

Using Machine Learning to Find Payment Anomalies

Next
Next

AI-Powered Trademark Search