辽宁大学学报（自然科学版）2023，Vol.50Issue(4)：302-311,10.
基于知识图谱的短视频实时推荐方法研究
Research on Real-Time Recommendation Method for Short Video Based on Knowledge Graph
摘要
Abstract
Short video enables users to obtain rich information in fragmented time,and it can be simply recorded,propagated at a high speed,used less bandwidth,so it has been favored by more and more people.But at the same time,the interests of short video users are time sensitive.Therefore,improving the real-time performance of short video recommendation is essential in current research.Knowledge graph(KG)can integrate the rich interactive relationship information and the attribute information of users and short video,so we use KG for representation.In order to improve the real-time performance of short video recommendation,this paper proposes a short video real-time recommendation method based on KG.Firstly,we extract the attribute features of short video through the attention mechanism,and then use GCN(Graph convolution network)with time information to represent the rich information of users and short video for short-term interest.At the same time,the historical browsing information is obtained from the graph convolution.Finally,the long-term interest and short-term interest are fused through the RNN(Recurrent neural network)model to get the final real-time recommendation results.Experiments show that the proposed method improves the precision and recall rate compared with the mainstream dynamic recommendation methods such as FM(Factorization machine)and LSTM(Long short term memory)network.关键词
实时推荐/短视频/知识图谱/图卷积网络/长短时记忆网络Key words
real-time recommendation/short video/knowledge graph/GCN(Graph convolution network)/LSTM(Long short term memory)network分类
信息技术与安全科学引用本文复制引用
冯勇,孙宇,徐红艳,王嵘冰..基于知识图谱的短视频实时推荐方法研究[J].辽宁大学学报（自然科学版）,2023,50(4):302-311,10.基金项目
辽宁省教育厅科学研究基金面上项目(LJKZ0085,LJKMZ20220447) （LJKZ0085,LJKMZ20220447）
辽宁省科学技术厅项目(2023JH4/10700056) （2023JH4/10700056）
2022年辽宁省本科教学改革项目(2022-21) （2022-21）
