NTNU helps K-12 students master English faster with innovative learning platform powered by Azure OpenAI Service

Preparing students for global success

According to a survey by the Taiwanese government, students in Taiwan have stronger skills in reading and writing English than in speaking the language. With an aim to empower Taiwanese students to improve English-language speaking skills to excel in all areas of industry on a global level, the Taiwan Ministry of Education worked with National Taiwan Normal University (NTNU) to launch “Cool English”—one of the world's first English-language learning platform of Microsoft Azure OpenAI Service.

The Cool English platform is designed for self-learning that also uses Azure Cognitive Service for Speech. Students gain skills through creative learning scenarios such as finishing song lyrics and problem solving while watching animated videos. Cool English bots provide human-like conversation and interaction, giving learners immediate feedback and highly realistic speaking training. Launched at the end of 2015, Cool English had racked up over 1.4 million registered users by 2022, mostly comprised of elementary and middle school students. 

Harnessing the power of Azure OpenAI Service 

Utilizing the Azure OpenAI Service advanced language models called Davinci and ChatGPT, NTNU built 10 educational “CoolE” bots for the Cool English platform, each with its own unique personality. NTNU designed the bots to engage students in various use cases, such as solving mysteries as a detective or gaining career insights from an enthusiastic data analyst. Their language is also adjustable to different ages and proficiency levels, providing a more personalized learning experience. 

“We needed something that could adjust its language based on each individual student’s needs and English proficiencies. The CoolE bots use simpler words and sentences, if necessary, which is a huge win for our students from all different backgrounds,” says Dr. Hao-Jan Howard Chen, Professor, Department of English at National Taiwan Normal University.

Compared to other AI models that NTNU used previously, the Davinci model in Azure OpenAI Service produces significantly more inviting and approachable responses. In choosing Microsoft, NTNU also uses voice recognition and pronunciation technology, which can grade the quality of students’ spoken English within the Cool English platform. 

“The bots we experimented with before were fairly robotic, but with the advanced language models in Azure OpenAI Service, the responses are much more inviting, approachable, and realistic, which makes a big difference for our students,” says Chen. Chen adds that the more realistic conversational practices foster more active learning, as opposed to passive methods like listening to movies or reading books in English. 

Paving the way for future AI use cases 

In the future, the Taiwan Ministry of Education and National Taiwan Normal University plan to use Azure OpenAI Service to further optimize prototype bots and promote Cool English extensively to teachers in Taiwan. NTNU is optimistic about what this sort of technology can mean for English language learning in the country, and how receptive teachers will be. 

“Teachers are already looking for tools like this that can better help students on speaking and writing in the future. As we look to market this to a larger audience of educators, we expect it to be well-received with little to no promotion, as the need is already there and strong,” says Dr. Chen.

NTNU is also planning to use other models in Azure OpenAI Service to build new features for Cool English that will help students with their writing. For example, NTNU has prototyped a product that will allow students to paste in an essay written in English and get immediate grammatical corrections and guidance in return.  

“By utilizing the advanced language models in Azure OpenAI Service, Cool English is taking an innovative step for English education in Taiwan, helping students reach their learning goals and overcome the challenges of limited opportunities for real-life conversational and writing practice,” says Dr. Chen.