Automated Analysis of Chemistry Experiment Videos: New Challenges for Video Understanding

Zhichao Zheng, Benhua Wang, Ziwen Wang, Yi Chen, Junsheng Zhou, Li Kong

Abstract: Compared to static declarative knowledge, procedural knowledge is challenging to assess effectively in education due to its nature of dynamic and complex. However, it serves as a crucial source for essential abilities of students. Can artificial intelligence assist in evaluating procedural knowledge? To explore this question, we focus on the scenario of middle-school chemistry experiments and attempt to use video understanding technology to aid teachers in assessing procedural knowledge of chemistry experiments. Nevertheless, our preliminary findings reveal that chemistry experiment videos differ from typical instructional videos used in research, presenting unique characteristics and complexities. Thus, we pose a new challenge, offering novel research questions for the field of video understanding and a new perspective for leveraging artificial intelligence in modern education.