2020/01/19 by Yueji Yang, Anthony K. H. Tung, Yang, Yueji +1
Computer Science · #Advanced Database Systems and Queries #Algorithms and Data Compression #Data Management and Algorithms #Databases (cs.DB) #FOS: Computer and information sciences #cs.DB
paper · pdf · doi:10.48550/arxiv.2001.06770
arxiv created 2020/01/19 · openalex publication_date 2020/01/19 · arxiv updated 2020/01/22 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Recently, keyword search on Knowledge Graphs (KGs) becomes popular. Typical keyword search approaches aim at finding a concise subgraph from a KG, which can reflect a close relationship among all input keywords. The connection paths between keywords are selected in a way that leads to a result subgraph with a better semantic score. However, such a result may not meet user information need because it relies on the scoring function to decide what keywords to link closer. Therefore, such a result may miss close connections among some keywords on which users intend to focus. In this paper, we propose a parallel keyword search engine, called RAKS. It allows users to specify a query as two sets of keywords, namely central keywords and marginal keywords. Specifically, central keywords are those keywords on which users focus more. Their relationships are desired in the results. Marginal keywords are those less focused keywords. Their connections to the central keywords are desired. In addition, they provide additional information that helps discover better results in terms of user intents. To improve the efficiency, we propose novel weighting and scoring schemes that boost the parallel execution during search while retrieving semantically relevant results. We conduct extensive experiments to validate that RAKS can work efficiently and effectively on open KGs with large size and variety.