DATA
BANK

DataBank's AI grading system employs a sophisticated learning approach that trains AI to learn the unique grading patterns of well-trained human graders, which is a more complex training strategy than similar services that train AI to simply learn the average results given by two or more graders. Additionally, by utilizing a relevance assessment module, our system can detect anomalies that other services often miss. It can also assess how well a student performs on a test based on the established grading criteria beyond basic syntax evaluation.

Databank leverages source data from over 200 countries including India and China, to manage key evaluation metrics for both native and non-native English STT(speech-to-tet) performance. As a result, our STT recognition system for non-native speech shows superior performance to OpenAI's Whisper.

Databank's AI-powered correction system is a proprietary generative model that is trained based on the grading criteria of various English tests. It goes beyond correcting the grammatical errors and incorrect expressions that ChatGPT can also modify, and generates revised versions that can achieve high scores based on the required grading criteria.