Exploiting anonymous entity mentions for named entity linking

被引:3
作者
Hou, Feng [1 ]
Wang, Ruili [1 ]
Ng, See-Kiong [2 ]
Witbrock, Michael [3 ]
Zhu, Fangyi [2 ]
Jia, Xiaoyun [4 ,5 ]
机构
[1] Massey Univ, Sch Math & Computat Sci, Palmerston North, New Zealand
[2] Natl Univ Singapore, Inst Data Sci, Queenstown, Singapore
[3] Univ Auckland, Sch Comp Sci, Auckland, New Zealand
[4] Shandong Univ, Inst Governance, Jinan, Peoples R China
[5] Shandong Univ, Sch Polit & Publ Adm, Jinan, Peoples R China
关键词
Entity linking; Fine-grained entity types; Anonymous entity type words; Entity embeddings;
D O I
10.1007/s10115-022-01793-3
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Named entity linking or named entity disambiguation is to link entity mentions to corresponding entities in a knowledge base for resolving the ambiguity of entity mentions. Recently, collective linking methods exploit document-level coherence of the referenced entities by computing a pairwise score between candidates of a pair of named entity mentions (e.g., Raytheon and Boeing) in a document. However, in a document, named entity mentions are significantly less frequent than anonymous entity mentions (e.g., defense contractor and the company). In this paper, we propose a method, DOCument-level Anonymous Entity Type words relatedness (DOC-AET), to exploit the document-level coherence between candidate entities and anonymous entity mentions. We use the anonymous entity type (AET) words to extract anonymous entity mentions. We learn embeddings of AET words from their inter-paragraph co-occurrence matrix; thus, the document-level entity-type relatedness is encoded in the AET word embeddings. Then, we compute the coherence scores between candidate entities and anonymous entity mentions using the AET entity embeddings and document context embeddings. By incorporating such coherence scores for candidates ranking, DOC-AET has achieved new state-of-the-art results on two of the five out-domain test sets for named entity linking.
引用
收藏
页码:1221 / 1242
页数:22
相关论文
共 50 条
  • [31] Entity linking for biomedical literature
    Jin G Zheng
    Daniel Howsmon
    Boliang Zhang
    Juergen Hahn
    Deborah McGuinness
    James Hendler
    Heng Ji
    BMC Medical Informatics and Decision Making, 15
  • [32] Language Independent Entity Linking
    Ding, Lei
    Dong, Bin
    PROCEEDINGS OF 2018 5TH IEEE INTERNATIONAL CONFERENCE ON CLOUD COMPUTING AND INTELLIGENCE SYSTEMS (CCIS), 2018, : 724 - 729
  • [33] Reddit entity linking dataset
    Botzer, Nicholas
    Ding, Yifan
    Weninger, Tim
    INFORMATION PROCESSING & MANAGEMENT, 2021, 58 (03)
  • [34] Entity linking for biomedical literature
    Zheng, Jin G.
    Howsmon, Daniel
    Zhang, Boliang
    Hahn, Juergen
    McGuinness, Deborah
    Hendler, James
    Ji, Heng
    BMC MEDICAL INFORMATICS AND DECISION MAKING, 2015, 15
  • [35] Towards better entity linking
    Mingyang LI
    Yuqing XING
    Fang KONG
    Guodong ZHOU
    Frontiers of Computer Science, 2022, 16 (02) : 56 - 68
  • [36] Towards better entity linking
    Mingyang Li
    Yuqing Xing
    Fang Kong
    Guodong Zhou
    Frontiers of Computer Science, 2022, 16
  • [37] Implicit Entity Linking in Tweets
    Perera, Sujan
    Mendes, Pablo N.
    Alex, Adarsh
    Sheth, Amit P.
    Thirunarayan, Krishnaprasad
    SEMANTIC WEB: LATEST ADVANCES AND NEW DOMAINS, 2016, 9678 : 118 - 132
  • [38] Entity Linking with Distributional Semantics
    Gamallo, Pablo
    Garcia, Marcos
    COMPUTATIONAL PROCESSING OF THE PORTUGUESE LANGUAGE (PROPOR 2016), 2016, 9727 : 177 - 188
  • [39] Entity Linking on Graph Data
    Yu, Minghe
    Feng, Jianhua
    WWW'14 COMPANION: PROCEEDINGS OF THE 23RD INTERNATIONAL CONFERENCE ON WORLD WIDE WEB, 2014, : 21 - 25
  • [40] Entity Linking in Queries Using Word, Mention and Entity Joint Embedding
    Wang, Zhichun
    Wang, Rongyu
    Wen, Danlu
    Huang, Yong
    Li, Chu
    SEMANTIC TECHNOLOGY, JIST 2017, 2017, 10675 : 138 - 150