The AI algorithm has been developed and trained on millions of data points over 8 years, using machine learning modelling, manual calibration of results and continual user testing. The algorithm is adaptive, continuously evolving and enhancing its performance.
In 2024, testing experts from telc ran a calibration event for Dynamic Speaking Test. See the report. Responses at different levels were evaluated by a panel of experienced human raters. The conclusion was: ‘The scores given by the group of experts aligned closely with those generated by the AI system. There was a high degree of agreement between human evaluators and the AI technology.’
At the end of the test, the user receives a detailed report containing: