CONFERENCE (INTERNATIONAL) Related Entity Expansion and Ranking Using Knowledge Graph
赤瀬 龍也, 河端 洋人, 西田 成臣, 田中 祐生, 神永 環
The 15th International Conference on Complex, Intelligent, and Software Intensive Systems (CISIS 2021)
July 01, 2021
Nowadays, it is possible for web search users to receive relevant answers, such as a summary and various facts of entity, quickly from a knowledge panel. Search services recommend relevant entities based on search trends. If the search engine recommends sufficient related entities, users will acquire adequate information of interest. This enhances user experience and for web service providers, increases the opportunity to attract users. In this study, we increase the number of knowledge panels that recommend related entities and optimize their order using a knowledge graph. We also introduce a production-level system that generates related entities from a massive knowledge base and search log; it achieves low latency serving. We deploy our system to the production environment and perform quantitative and qualitative estimation using A/B testing. Based on the results, we conclude that our method significantly enhances the impression based coverage of knowledge panels preventing a significant change in click-through rate.
Paper : Related Entity Expansion and Ranking Using Knowledge Graph (external link)