TAL Education GroupAutomated grading and retrieval

One of Asia's leading education technology companies processes hundreds of thousands of new questions and answers daily across digital classrooms serving millions of students.

Teachers spent 15–20 minutes manually evaluating single tests, while subjective questions often suffered from inconsistent scoring across different graders. Traditional fuzzy matching systems could not understand the semantic meaning of student responses or scale to support real-time automated feedback.

The organization implemented a vector retrieval platform using Milvus to manage over one billion vectors, utilizing MySQL to synchronize metadata with unstructured content. A dedicated pipeline processes multimodal inputs like handwritten formulas via OCR before generating embeddings for semantic search. A closed-loop quality framework continuously refines the grading logic by incorporating confidence scores and teacher feedback.