面向特定科研任务的著者姓名消歧方法

Method for Author Name Disambiguation in Specific Research Tasks

人才流动、学者评价等以学者个人为对象的研究任务,通常需要针对学术论著数据集中的著者进行姓名消歧。本文针对此类特定研究任务,提出了准确且便于学者操作的姓名消歧方法。为简便计算,弥补本地数据缺失的问题,本文构建了基于异源数据的二阶段姓名消歧框架。一阶段充分挖掘本地关联信息,二阶段结合权威的外源数据。基于表征进行本地关系发现、半模糊检索等步骤,以达到全面客观的姓名消歧,最终通过人工智能领域的论文数据和Aminer姓名消歧数据集,实现并验证该方法的优越性和普适性。经过与人工标注数据对比,该框架表现出良好的消歧效果,较好地解决了原始数据中的同名异人和同人异名问题,从而为后续研究任务奠定了扎实的基础。

Author name disambiguation is usually required in analyzing the flow of talents and evaluating scholars in aca‐demic works.This paper proposes an accurate and convenient method for name disambiguation for a specific research task.In order to simplify calculations and account for the lack of local data,this paper constructs a two-stage name disam‐biguation framework based on heterogeneous data.The first stage involves fully mining the local associated data,and the second stage combines the authoritative external data.Based on representation,relevant information extraction,relational network construction,semi-fuzzy retrieval,and other steps are carried out to achieve comprehensive and objective name disambiguation.Finally,the superiority of this method is identified and verified through thesis data under the field of artifi‐cial intelligence.Compared with manually annotated data,the framework performs better in disambiguation,and solves the problem of synonyms and namesakes in the original data,thus laying a solid foundation for subsequent research tasks.