📧 Email: [email protected], [email protected]
Google Scholar, DBLP
🎓 Education
Ph.D. in Computer Science and Engineering (On Going)
University of Notre Dame, Notre Dame, IN, US
M.Sc. in Computer Science and Technology ****
Shandong University, Qingdao, Shandong, China
B.Eng. in Electronic Science and Technology
“Chongxin Class”, Shandong University, Qingdao, Shandong, China
GPA: 87.02/100
Rank: 4/12
08/2023 - Current
Supervised by Prof. Meng Jiang
09/2020 - 06/2023
Supervised by Prof. Liqiang Nie
09/2016 - 06/2020
📜 Publications
Multimodal Activation: Awakening Dialog Robots without Wake Words
Liqiang Nie, Mengzhao Jia, Xuemeng Song, Ganglu Wu, Harry Cheng and Jian Gu. (SIGIR 2021)
- Define a new task that targets at using multimodal signs to awaken dialog robots without wake words.
- Divide multimodal activation task into two key sub-problems, i.e., audio-visual consistency detection and semantic talking intention inference.
⚒️ Intern & Activities Experience
R & D Intern, Kuaishou Technology, Beijing, China
04/2021 - 01/2022
- Devoted to the task of search oriented multimodal micro-video captioning. Used PLMs for improving evaluation performances. Designed a model to adapt data in the micro-video domain.
- Investigated existing large-scale multimodal pre-training tasks. To improve the effectiveness of real-life applications, we collected millions of micro-video data for pre-training.
Algorithm Intern, Momenta (Suzhou) Limited
01/2020 - 09/2020
- Adopted style transfer algorithm to remove the camera ISP module and adapt RAW format data directly to object detection task. Successfully implemented the use of RAW data to improve detection accuracy.