Abstract: The quality of teacher-child interaction is one of the most important process quality indicators of preschool education quality,and it is also a difficult point to break through in scientific and large-scale evaluation.Automated classroom observation is a new paradigm of multi-modal data-driven education research,which provides a new direction for teacher-child interactions quality assessment.Automated classroom observation has covered many areas of teacher-child interactions,such as emotional support,classroom organization and instruction support,among which the evaluation of emotional support is the most mature.Automated assessment also uses facial expression,voice,eye movement and other data modes,and voice data is the most widely used at present.Automated classroom observation can help to achieve integrated teaching and evaluation and large-scale evaluation,but it still faces challenges in data collection,quality and ethics.In the future,China can focus on five aspects,such as formulating evaluation indicators,focusing on breakthroughs in key technologies,building video resource database,carrying out interdisciplinary research and improving norms and standards,and driving the reform process of intelligent evaluation of preschool education quality with the application of automated classroom observation.

Key words: quality of teacher-child interactions, automated assessment, classroom observation, multi-modal data