Tsinghua Research Team Honored at 47th International ACM SIGIR Conference
The 47th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR 2024) was held in Washington, D.C., USA, from July 13 to 18. Faculty members and students from the Information Retrieval Lab (THUIR), Department of Computer Science at Tsinghua University received multiple awards at the conference.
The paper titled Explicit Factor Models for Explainable Recommendation based on Phrase-level Sentiment Analysis, published in 2014 by Yongfeng Zhang (PhD student) and Guokun Lai (undergraduate student) from THUIR and Dept. CS&T, supervised by Professor Min Zhang, Professor Yiqun Liu, and Professor Shaoping Ma, received the sole Test of Time Award in ACM SIGIR 2024. This paper is the first study defining the problem of explainable recommendation, and has developed effective sentiment analysis algorithms to address this technical challenge. With nearly a thousand citations, the paper has continuously played a pivotal role in research and applications in recommendation systems since its publication. The Test of Time Award is selected by the SIGIR committee annually to academic papers published at the SIGIR conference over ten years ago to acknowledge their significant and enduring impact to the field. This year marked the first time that a research institution from the Chinese mainland has received the honor.
The paper titled Scaling Laws for Dense Retrieval, authored by Yan Fang (MS student), Jingtao Zhan (Ph.D. student) and others from THUIR and Dept. CS&T, supervised by Assistant Professor Qingyao Ai and Professor Yiqun Liu, received the SIGIR’24 Best Paper Award. It investigates the scaling laws in dense retrieval models, an area not fully explored compared to language generation. By implementing models with varying parameters, training them with different amounts of annotated data, and introducing a continuous metric for performance evaluation, this paper reveals that dense retrieval models’ performance adheres to a power-law scaling with respect to model size and annotation quantity across various datasets and methods. These findings help optimize training processes, particularly under budget constraints, and provide valuable insights for future design of information retrieval systems such as search engines and recommendation systems. This is the first time the award has gone to a research institution from the Chinese mainland.
Assistant Professor Qingyao Ai from THUIR, Dept. CS&T has been selected to receive the SIGIR Early Career Researcher (ECR) Award for Excellence in Research. This award, accompanied by a plaque and honorarium, is exclusively granted to young scholars who have received their doctoral degrees within the past seven years. It acknowledges their significant impact on information retrieval research, community engagement, or DEI (i.e., Diversity, Equity, and Inclusion). The SIGIR ECR award for Excellence in Research also marks the first time a researcher from the Chinese mainland has received this honor.
Information Retrieval is a research field focused on collecting, processing, organizing, and storing information for retrieving and utilizing information based on user needs. It belongs to interdisciplinary research fields such as computer science, information science, cognitive psychology, etc. Technologies derived from information retrieval, such as search engines, recommendation systems, and conversational systems, have become the foundations of information societies.
The ACM Special Interest Group on Information Retrieval (ACM SIGIR) is the most famous international academic organization in the field of information retrieval. It hosts prestigious international conferences such as SIGIR, CIKM, and WSDM, and is widely recognized in both academic and industrial communities. Notably, the International ACM SIGIR Conference on Research and Development in Information Retrieval has a history of 47 years, serving as the primary platform for researchers worldwide to present cutting-edge research in this field. It is also recognized as an A-level conference by both the China Computer Federation (CCF) and the Chinese Association for Artificial Intelligence (CAAI).