|国家自然科学基金||Reasoning Methods Based on Graph Neural Networks for Personalized Recommendation, No. U19A2079||2020.1-2023.12||RMB 2,410,000||国家自然科学基金重点项目|
|国家自然科学基金||Research on Conversational Recommendation, No. 61972372||2020.1-2023.12||RMB 600,000||国家自然科学基金面上项目|
Reasoning Methods Based on Graph Neural Networks for Personalized Recommendation
This project aims at constructing effective recommendation models that can provide the reasons behind results:
1. From the perspective of graph reasoning, we will integrate the technologies from information retrieval, machine learning and network science, focusing on the key scientific problems of large-scale graph representation learning, context-aware user long-term and short-term interest representation, coupled learning of data and knowledge, and exploratory user interest reasoning.
2. We will provide an in-depth study of reasoning methods based on graph neural network, and conduct research on reasoning over knowledge graph, social network, spatio-temporal context and interactive conversations for recommendation.
3. Finally, we will form a set of recommendation theory and technical system based on graph reasoning, leading the next generation of recommendation technology moving towards the direction of knowledge value-added services.
Research on Conversational Recommendation
This project aims to explore personalized recommendation methods in the conversational setting, so as to provide the research foundation for the next generation of recommender systems:
1. The core scientific questions in multi-round conversational recommender system are three-fold: 1) how to predict user preferences from historical behaviors. 2) how to develop optimal strategies for recommending items and asking questions across the sessions. 3) how to update the recommender model and the conversation strategy based on the real-time feedback from users.
2. Focusing on the three scientific questions, we will carry out a series of research aimed at providing a complete set of solutions for constructing an effective and practical conversational recommender system.
3. Through these studies, we hope to get achievements in the international communities of information retrieval and recommender system, promoting the application of conversational recommender systems in the Internet industry and the development of Web Intelligence in China.