基于微博文本的层次化实体链接方法-yuhengli.pdfVIP

基于微博文本的层次化实体链接方法 1,2,3 1,2,3 1,2,3 1,2 2 1,2,4 李禹恒 ；宋俊 ；黄宇 ；付琨 ；吴一戎 ；陈昊 （1.中国科学院空间信息处理与应用系统技术重点实验室，北京 100190；2.中国科学院电子学研究所， 北京 100190；3. 中国科学院大学，北京 100049；4.北京空间信息中继传输技术研究中心，北京 100094） 摘要：命名实体链接是自然语言理解的重要研究内容，同时也是知识图谱构建及实体搜索的基础。本文出一种针对微博文 本的层次化实体连接方法 （HEL ）。基于用户偏好一致性假设，该方法首先对所有及根据信息函数进行排序，得到歧义最 小的及利用消歧算法消歧，并将返回的确认实体包含进消歧函数。通过这种迭代策略让正确的结果正向传递给下一层更模 糊的消歧任务。在人工标注测试集上的实验表明，本文的方法表现出良好的性能。 关键词：实体链接 文本消岐 数据挖掘 中图分类号：G202 Hierarchical Entity Linking based on Microblogs LI Yu-heng, SONG Jun, HUANG Yu, FU Kun, WU Yi-rong, CHEN Hao (1. Key Laboratory of Technology in Geo-spatial Information Processing and Application System, Institute of Electronics, Chinese Academy of Sciences, Beijing 100190; 2. Institute of Electronics, Chinese Academy of Sciences, Beijing 100190; 3. University of Chinese Academy of Sciences, Beijing 100 049; 4. Beijing Space Information Relay Transmission Technology Research Center, Beijing 100094 ） Abstract ：Named Entity Linking is a crucial research of Nature Language Understanding, and the foundation of Knowledge Graph construction and Entity Searching. This paper provides a Hierarchical Entity Linking method (HEL) based on Twitter content. Considering the assumption of user preference consistency, the method first rank all the candidates mentions based on proposed Information Function, and then assign the most familiar candidate to the given mention by adopting a Scoring Function. This procedure will iterate by incorporating disambiguated entities into the Scoring Function, which consequently passes on the certainty from previous linking results to the following rounds of m