摘要
针对现有基于内容的推荐方法中存在的知识利用不充分问题,提出了一种融合关系抽取的推荐系统,在用word2vec模型对物品知识进行编码的基础上,用补充模板特征的关系抽取模型对物品知识进行了更深层次的挖掘,构建了增强知识图谱,进而获得增强实体特征,与文本特征、基础实体特征融合后构建物品特征.实验证明,融合关系抽取的推荐系统推荐效果优于同类模型,并且各个部分的改进都是有效的.
Aiming at the problem of insufficient knowledge utilization in the existing content-based recommendation methods,a recommendation system based on fusion relation extraction was proposed in this paper.Using word2vec model to encode object knowledge,using supplementary template features to excavate the object knowledge in a deeper level,an enhanced knowledge graph was constructed.Moreover the enhanced entity features were obtained,being combined with text features and basic entity features to construct object features.Experimental results show that the recommendation effect based on fusion relation extraction is better than that of the similar models,and the improvement of each part is effective.
作者
高春晓
卢士帅
刘琼昕
宋祥
GAO Chunxiao;LU Shishuai;LIU Qiongxin;SONG Xiang(Beijing Engineering Applications Research Center on High Volume Language Information Processing and Cloud Computing,Beijing Institue of Technology,Beijing 100081,China;School of Computer Science and Technology,Beijing Institue of Technology,Beijing 100081,China)
出处
《北京理工大学学报》
EI
CAS
CSCD
北大核心
2022年第11期1191-1199,共9页
Transactions of Beijing Institute of Technology
基金
国家重点研发计划项目(2020AAA0104903)
国家自然科学基金资助项目(62072039)。
关键词
人工智能
深度学习
关系抽取
推荐系统
模板特征
artificial intelligence
deep learning
relational extraction
recommendation system
template features
