Recently, the paper "Representation and Extraction of Physics Knowledge Based on Knowledge Graph and Embedding-Combined Text Classification for Cooperative Learning" co-first authored by SHANG Jialin, HUANG Jingyuan and ZENG Shihua, undergraduates of class of 2024, majoring in computer engineering, were accepted by the 25th IEEE CSCWD International Conference. The conference is one of the most important international conferences in the field of collaborative computing, and it is also a key conference recommended by CCF of China computer society.
▲ From Left: HUANG Jingyuan, ZHANG Jian, WANG Hongwei, SHANG Jialin, ZENG Shihua
It is worth mentioning that the main work of the paper was completed by three students in their second semester of freshman year, and the following summer vacation. It is an important achievement of their freshman scientific research and training project (SRTP) "Research on Intelligent Evaluation System Of Subject Knowledge Level Based On Knowledge Graph". The project is jointly guided by ZJUI Professor WANG Hongwei and Professor Wang's Research Assistant ZHANG Jian. This paper builds an embedded layer combined neural network for high school knowledge text classification, and adds the classification results to the knowledge graph for combing knowledge and visualizing the output of knowledge, which improves the efficiency of neural network for physical knowledge text classification, and realizes the data and systematic management and integration of high school physical knowledge by using neural network system and knowledge graph. It expands the application scope of knowledge atlas and neural network collaborative learning, as well as the application scope of neural network in word embedding and text classification.
Research journey started from the freshmen year
HUANG Jingyuan participated Professor WANG's lecture in a Residential College Tutor salon. The lecture not only inspired him deeply, but also ignited his interest in the research direction of knowledge engineering. Jingyuan participated in the information competition in high school and also conceived how to build a topic recommendation system to provide personalized learning programs to help students learn better in both competition and the college entrance examination. After listening to Professor WANG's lecture, it coincided with that ZJUI encouraged undergraduates to apply for SRTP project. HUANG Jingyuan took the initiative to consult Professor WANG on the feasible scheme of building the system, hoping to broaden their thinking of data analysis by building the system. He also invited SHANG Jialin, who has experience in subject research and physics competition training, and ZENG Shihua, who has experience in physics competition training and artificial intelligence winter camp, to participate in the subject and successfully won the project. Based on their common interest in the knowledge graph and in-depth learning direction of artificial intelligence, the three like-minded friends started their research road with enthusiasm and interest in scientific exploration. Under the guidance of their instructor, the three students finally decided to focus the project on the construction of high school subject knowledge intelligent evaluation system based on the construction of knowledge graph and neural network classification.
▲ Online Group Discussion
After the project was selected as the university level SRTP project, they, even though all freshmen, began to join Professor WANG's laboratory and start doing research for real. They made full use of their summer vacation and after-school time, worked hard, and carried out a lot of research work. Professor WANG and his research assistant ZHANG Jian also gave a lot of help and guidance to this project. Taking advantage of the group meeting and the opportunity of face-to-face exchange and learning, they kept in-depth discussions with their instructors, asked about the construction of the project framework, and therefore learned the specific application of artificial intelligence in the project and the basic skills of research.
▲ The iterative running results of the programs written by three students
Being able to join professor's research group from freshman year, students benefit from ZJUI's extensive encouragement and active mobilization of undergraduates to carry out research projects, as well as the SRTP project for "scientific research green hand". By encouraging students to participate in the SRTP project, we can help undergraduates with little experience get a real opportunity to enter the laboratory, learn from professors, and drive ZJUI to form an excellent learning atmosphere and scientific research atmosphere. These efforts fully reflect ZJUI's great attention to the cultivation of undergraduates' innovative thinking and research practice ability. "In addition to improving the quality of scientific research, this unforgettable SRTP has also helped us broaden our vision, learn how to carry out teamwork, and played a positive role in exploring the future research areas and development path. " HUANG Jingyuan said.
Using computer collaborative learning to serve the cause of education
Artificial intelligence is the key field of new infrastructure, and knowledge graph is the bottom support of cognitive intelligence. The three students focused on the most cutting-edge disciplines and comprehensively applied the hot technologies of academic and industrial, such as web crawler, neural network and deep learning, knowledge graph, recommendation system and so on.
▲ Basic Structure of Embedding Layer Combined Neural Network
▲ Use knowledge graph to sort out high school physics knowledge
The intelligent system for systematic management and integration of high school physics knowledge proposed by the project aims to solve two major problems existing in current Chinese high school education: one is the problem of online learning resources for Chinese high school students. At present, the types of online learning resources for senior high school students are complex and the quality is uneven. They are often difficult for high school students to select. They may waste a lot of time and energy in screening resources, resulting in low learning efficiency; Second, the traditional teaching in senior high school is not personalized and targeted enough. Because there are great differences between students' knowledge level and learning progress, it is usually difficult for teachers to make an overall plan. HUANG Jingyuan, SHANG Jialin and ZENG Shihua put forward the idea of better integrating the traditional education industry with information technology, which is conducive to the systematic evaluation of students' subject knowledge and ability level, and also provides many conveniences for teachers to master and analyze students' learning more efficiently. At the same time, it enriches the ways for students to obtain learning resources, so that students (especially students in areas with relatively scarce educational resources) can obtain many rich learning resources online that they could not obtain before. Through this project, the three students hope to optimize high school education resources, help high school students formulate learning strategies scientifically, help middle school teachers provide more personalized guidance, use the scientific and technological dividends of the information age to reduce the burden on teachers, and use computer-based collaborative learning to serve the education industry.
▲ The text classification model constructed with ERNIE+BERT as the embedding layer and RCNN as the upper neural network can achieve a classification accuracy of 94.39%
IEEE CSCWD is one of the most important international academic conferences in the field of collaborative computing. It is a key conference recommended by China Computer Federation(CCF). It has been the 25th conference by 2022. Scholars at home and abroad from more than 20 countries and regions participate in the conference every year. The conference admission papers will be published in the IEEE Classification Conference Papers Collection and will be submitted to EI, DBLP, etc. for indexing. The acceptance of the sophomores’ papers by the high-level international conference, fully shows that the students have solid professional knowledge, language ability and excellent scientific research exploration and knowledge application ability, and highlights the unique advantages of ZJUI's international cooperative running mode in internationalization and interdisciplinary education.
Group Overview
The project relies on the research team of knowledge engineering and knowledge system led by Professor WANG Hongwei. The team mainly focuses on the integration of artificial intelligence technology and industrial system, including the research work of knowledge engineering and system, industrial knowledge atlas, data-driven fault analysis and diagnosis, and have established in-depth cooperation with Cambridge University, UIUC, Tsinghua University, North China Electric Power University, and other universities at home and abroad, which fully reflects the scientific research characteristics of international cooperation and interdisciplinary in the International Campus of Zhejiang University. Although it has been established for a short time, at present, the team has more than 20 researchers, who undertake important projects such as national key plan, natural science foundation and Zhejiang University-UIUC joint fund. The corresponding achievements provide technical support for intelligent applications of large enterprises such as China nuclear power and China Aerospace. After several years of efforts, the team has accumulated rich achievements. Since January 2022, many papers have been approved by IEEE Trans. on Industrial Informatics, IEEE Trans. on Instrumentation and Measurement, Applied Soft Computing and other high-level journals.
