Recently, the TAL AI Engineering Institute's machine learning team had seven academic papers consecutively accepted into several top international academic conferences, including the International Conference on Artificial Intelligence in Education (AIED 2020), the International Conference on Educational Data Mining (EDM 2020), the International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2020), and the World Wide Web Conference (WWW 2020). These papers showcase the potential of AI+Education development from China to the global audience.

The accepted papers primarily focus on research in AI+Education application scenarios, covering multiple branches of artificial intelligence studies such as speech recognition, data mining, and machine learning. Three papers were accepted into the AIED 2020 conference.

AIED is renowned as a top international conference in the educational application field, known for "providing high-quality research of intelligent systems and cognitive science methods in the educational computing application domain." Papers collected by AIED represent the latest developmental directions and levels of AI applications in the education field.

The three papers accepted by the AIED 2020 conference are as follows: "Siamese Neural Networks For Class Activity Detection," focusing on teacher voice recognition and separation, achieved AUC results of 94.2% and 85.5% in one-on-one online and offline small class scenarios respectively; "Neural Multi-Task Learning for Automatic Detection of Teacher Questions in Online Classrooms," which proposes a novel framework for automatically detecting teacher questions in online classes, quantifying teacher behavior through the detection of different types of questions such as open-ended, inquiry, dialogue management, and procedural questions; "Automatic Dialogic Instruction Detection for K-12 Online One-on-one Classes," targeting automatic detection of language behaviors related to teachers, provides teachers guidance on classroom behavior adjustments across different subjects and grades, helps improve teaching techniques and quality.

At the EDM 2020 conference, a leading international conference in education big data, TAL's paper "Identifying At-Risk K-12 Students In Multimodal Online Environments: A Machine Learning Approach" was successfully accepted. This marks the first attempt by industry and academia to predict dropout behavior in the K12 online education scene. By analyzing multidimensional data from classroom behavior and after-class services, it enables timely understanding of students' learning statuses and knowledge mastery conditions, assisting in optimizing student learning plans.

During the ICASSP conference, the world's largest and most comprehensive conference on signal processing and its applications, TAL's new model papers led to enthusiastic reactions. "Multimodal Learning For Classroom Activity Detection" presented a multimodal speaker recognition model method based on voiceprint attention structure, showing an accuracy improvement of about 10% over SOTA models, demonstrating the model's excellence in speaker segregation results within educational scenarios. In another paper, "UPGRADING CRFS TO JRFS AND ITS BENEFITS TO SEQUENCE MODELING AND LABELING," TAL upgraded the classic sequence model CRF to a joint generative model JRF, with new models consistently exceeding CRF in algorithm indicators, offering greater potential for improvements across various domains of sequence modeling and labeling tasks.

Moreover, TAL's paper on spoken language proficiency assessment in free scenarios, "Dolphin: A Spoken Language Proficiency Assessment System for Elementary Education," was also accepted into the top international internet conference WWW2020 and presented at the meeting. The paper, based on oral expression proficiency assessment innovations developed by TAL AI Engineering Institute, addresses the issue of inefficient, scalable, standardized evaluation of students' oral expression abilities.

It is noteworthy that more than 70% of technical staff from TAL AI Engineering Institute's machine learning team participated in publishing papers and patents. Recently, TAL also had multiple academic achievements accepted into AAAI 2020, NCME2020, and other top international academic conferences. TAL AI Engineering Institute won the EmotioNet facial expression recognition competition at the world's top conference in computer vision, CVPR2020.

Machine Learning Team of TAL  AI Lab

A series of AI research achievements being continuously accepted by top international academic conferences signifies the international academic community's recognition of Haoweilai's research strength and also indicates that Haoweilai is providing more practical value in educational scenarios through the use of AI technology.

Haoweilai's exploration and achievements in the field of AI research stem from its understanding and emphasis on AI+ education.

In recent years, Haoweilai has continuously increased its investment in AI research and has developed over 100 AI capabilities in 8 major types including image, speech, data mining, natural language processing, and more, around the needs of educational scenarios. It has also created over 10 AI application solutions covering various teaching processes. Currently, Haoweilai is also productizing multiple AI capabilities and widely applying them in multiple internal business operations.

Based on its internal successful empowerment, Haoweilai actively promotes the open sharing of AI capabilities across the industry and provides leading AI capabilities and solutions to industry partners through the AI Open Platform.

As of now, the AI Open Platform of Haoweilai has launched more than 100 AI capabilities for the education industry, with half of them being leading and unique models. In August of last year, the Ministry of Science and Technology approved the construction of the National New Generation of Artificial Intelligence Open Innovation Platform for Smart Education based on Haoweilai.

Looking to the future, Haoweilai will continue to adhere to the principles of original innovation, application-driven, and open sharing, take on the responsibility of promoting the integrated development of AI+ education, and work with industry partners to support the intelligent upgrade of the education industry.