基于多源知识的中文微博命名实体链接  

Chinese Micro-blog named entity linking based on multisource knowledge

作　　者：昝红英[1] 吴泳钢[1] 贾玉祥[1] 牛桂玲[2] 

机构地区：[1]郑州大学信息工程学院,河南郑州450001 [2]郑州大学外语学院,河南郑州450001

出　　处：《山东大学学报（理学版）》2015年第7期9-16,共8页Journal of Shandong University(Natural Science)

基　　金：国家自然科学基金资助项目(61402419;60970083;61272221);国家社会科学基金资助项目(14BYY096);国家高技术研究发展计划863计划项目(2012AA011101);河南省科技厅科技攻关计划资助项目(132102210407);河南省科技厅基础研究资助项目(142300410231;142300410308);河南省教育厅科学技术研究重点项目(12B520055;13B520381);计算语言学教育部重点实验室(北京大学)开放课题资助项目(201401)

摘　　要：命名实体在文本中是承载信息的重要单元,而微博作为一种分享简短实时信息的社交网络平台,其文本长度短、不规范,而且常有新词出现,这就需要对其命名实体进行准确的理解,以提高对文本信息的正确分析。提出了基于多源知识的中文微博命名实体链接,把同义词词典、百科资源等知识与词袋模型相结合实现命名实体的链接。在NLP&CC2013中文微博实体链接评测数据集进行了实验,获得微平均准确率为92.97%,与NLP&CC2013中文实体链接评测最好的评测结果相比,提高了两个百分点。Named entity is an important component conveying information in texts. Micro-blog is a social network platform used to share brief real-time information, with characteristics such as short text length, nonstandard words, and even the frequent emergence of neologisms. So an accurate understanding of the named entities is needed to ensure a correct analysis of the text information. A Chinese Micro-blog entity linking strategy was proposed based on multi-resource knowledge, combing the dictionary of synonyms, the encyclopedia resources as well as the bag-of-words model together to deal with named entity linking. In this strategy, named entities to be linked in Micro-blog were mapped to the corresponding candidate entities in the knowledge base. The evaluation results obtain a micro average accuracy of 92. 97%, based on experiments using data sets of NLP＆CC2013 Chinese micro-blog entity linking track. Compared with the state- of-the-art result, the accuracy of this method is two percent higher, which demonstrates the effectiveness of our method.

关 键 词：命名实体 中文微博实体链接 同义词词典 百科资源 词袋模型 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]