异构信息空间中时间感知的跨类型实体推荐

摘要

With entity search has become a new trend of information retrieval,entity recommendation has also become one of the hottest research problems in industry and academia.Due to heterogeneous entities with rich associations in heterogeneous information spaces,so cross-type entity recommendation is vital.Moreover heterogeneous entities have time information and evolve over time in heterogeneous information spaces,users want to get the most time relevant entity recommendation.In this paper,a time-aware cross-type entity recommendation framework T-ERe was proposed,which leverages rich associations among different entity types and query log to realize cross-type entity recommendation.T-ERe considers temporal information of entities and recommends the most time-relevant various types of entities to users.Experimental results on two real data sets demonstrate the feasibility and effectiveness of T-ERe.%随着实体搜索成为信息检索的一种新趋势,实体推荐也成为业界和学术界的热门研究问题之一.异构信息空间中的异构实体间彼此相互关联,因此跨类型实体推荐至关重要.此外,异构实体具有时间信息,异构信息空间中的实体不断随时间演化,用户希望得到在时间上最相关的实体推荐.提出一个时间感知的跨类型实体推荐框架T-ERe,利用异构实体间丰富的关联关系和查询日志实现跨类型的实体推荐.T-ERe考虑实体的时间信息和查询的时间上下文,给用户推荐时间上最相关的多种类型的实体.在真实数据集上的实验结果表明了T-ERe的可行性和有效性.