The Center for Artificial Intelligence, in collaboration with the School of Foreign Languages and the online campus of the National Research University Higher School of Economics (HSE University), has developed the "AI Lingvo" neural network. The development allows assessing written and oral English language skills. This was reported in the university's press service.
To test their knowledge, users are invited to write an essay, construct a monologue, and answer interview questions. The test can be taken on any device. The execution time is no more than 60 minutes, and the review with comments takes 30 minutes.
At the end of the testing, an overall score is determined, and a language proficiency level is assigned, from Elementary to Upper-Intermediate. The scale of assessment levels in the neural network correlates with all international examination systems.
Our development is distinguished by its ability to simultaneously assess not only text responses but also audio speech, and to evaluate the overall level of language proficiency based on a set of criteria. To do this, we needed to study a significant number of open solutions and, ultimately, conduct additional training on works checked by professional linguists.
The neural network is trained on the basis of thousands of expert assessments of oral and written texts. Currently, "AI Lingvo" is in the beta version stage. After calibration and testing, it is planned to be integrated into the learning process to evaluate students' work in English.
Earlier, www1.ru reported that 20 languages will appear in Yandex's "Translator" service.
Read materials on the topic:
Yandex's neural network has learned to follow the thread of the conversation
Yandex has launched a free interactive whiteboard service to replace the departed Miro
