Case Study: Astrid and Oxford Languages

Astrid & Oxford Languages

Affordable language learning for the next generation

The Product New


 

Astrid Education (“Astrid”) are an innovative educational technology company providing a voice-based, AI powered language learning platform. Language learning solutions are often expensive, and can be out of reach in emerging markets, or in places where there is a lack of English-speaking teachers. To bridge the language barrier, students are often forced to pay for costly private tutoring in order to achieve fluency.

 

With a focus on engaging learners through storytelling and conversation activities, the Astrid app offers an accessible, efficient, and affordable learning tool for children and teenagers who want to develop the practical side of their English skills, particularly speaking and pronunciation.

 

Astrid developed a speech recognition technology tailored to language learning. In the app a user reads out loud from a library of texts, including stories and exercises, and their speech is then analysed, and feedback is provided. The user journey is then adapted to the individual learner based on their Common European Framework of Reference (CEFR) level, with tailored suggestions made for improvement.

 

Communicating confidently in English can be key for career development and accessing information from around the world. Astrid’s app enables a communicative learning approach to supplement the education provided by teachers, schools, and parents. By making high-quality private English tutoring more affordable, Astrid enables students from around the world to compete in the global talent market.

 

 

The Problem


 

Astrid needed a partner to collaborate with to achieve their goal of developing world-leading speech recognition technology. Being new to the market, they required access to quality data and language expertise to enable fast development of their product, and to allow for continued evolution, pushing the boundaries of language learning solutions.

"Our research-focused partnership with Oxford Languages has enabled us to further develop our proprietary AI technology, to provide precise scoring and give correct feedback to empower the next generation of English learners."

 

— Vasilis Kalogiras, Senior Machine Learning Engineer at Astrid

 

Astrid partnered with Oxford Languages, undertaking a six-month research agreement using our flagship English language datasets, the Oxford Dictionary of English (ODE) and the New Oxford American Dictionary (NOAD), as well as audio content, to validate the AI that underpins their product.

 

The inclusion of using both ODE and NOAD datasets allows Astrid to build and curate content based on the International Phonetic Alphabet (IPA). Astrid’s team can look at the IPA for specific words in the dictionaries, and further research the possibility for users to be able to choose between learning British and American English.

 

Oxford Languages audio files are used to train the app’s machine learning algorithm that determines whether a user has pronounced a word correctly, with the algorithm providing a score for each pronunciation. Oxford Languages audio data provides the highly accurate data needed for more precise scorings.

 

The inclusion of country specific accents included in Oxford Languages audio data further trains and provides nuance to Astrid’s algorithm, enabling them to address localised learning challenges users may face.

 

Astrid’s collaboration with Oxford Languages provides a trusted partner embedded in the industry, and allows them to continue creating accessible language learning solutions.