Author Name Disambiguation in Medline

描述(由申请人提供)： 目前无法识别哪些论文具有相同的作者姓名(姓氏，首字母)是由不同的人撰写的，这阻碍了用户检索与健康相关的信息，以及致力于了解生物医学科学家的发表和合作行为的研究。消除作者姓名的歧义将有助于科学计量学和卫生政策研究，以及许多种类的日常科学任务：例如，选择裁判和与会者。我们已经创建了一个概率模型，说明Medline文章的属性在不同作者之间是如何变化的，并假设这可以作为消除Medline中作者姓名歧义的基础。在这项为期两年的探索性研究中，建议： 1.建立和评估“作者-个人”数据库，该数据库列出Medline中的所有论文，并将其中的绝大多数分配给一个或多个特定的作者--具有高度信心的个人。将改进一个基于Medline记录字段的概率模型，该模型估计任何两篇同名论文由同一人撰写的概率，包括补充信息，如所有作者的名字和从属关系。然后，对聚类算法进行优化和应用，为Medline中的所有名字形成作者-个人聚类。 2.更新作者-个人数据库(每周)和基本概率模型(每年)，并创建和评价一个免费的、公开的、多用户查询界面。该数据库还将提供给学术研究人员，用于文献计量学、科学计量学和政策研究。 这项研究将为未来更深入地研究出版和合作行为奠定基础，这将为提高生物医学科学的科学生产率提供有价值的见解。

DESCRIPTION (provided by applicant): The current inability to identify which papers bearing the same author name (last name, first initial) are written by different individuals is an impediment to user retrieval of health-related information as well as research devoted to understanding the publication and collaboration behavior of biomedical scientists. Disambiguation of author names will help in scientometrics and health policy studies, as well as everyday scientific tasks of numerous kinds: for example, choosing referees and conference attendees. We have created a probabilistic model of how the attributes of Medline articles vary across authors, and hypothesize that this can serve as the basis for disambiguating author names in Medline. In this exploratory two-year study, it is proposed: 1. To create and evaluate a database of "author-individuals" that lists all of the papers in Medline and assigns the great majority of them to one or more specific author-individuals with high confidence. A probabilistic model based on Medline record fields will be refined which estimates, for any two papers bearing the same name, the probability that they were written by the same individual, including supplementary information such as author first names and affiliations for all authors. Then, clustering algorithms will be optimized and applied to form author-individual clusters for all names in Medline. 2. To update the author-individual database (weekly) and underlying probabilistic model (yearly), and to create and evaluate a free, public, multi-user query interface. The database will also be made available to academic researchers for bibliometric, scientometric and policy studies. This research will set the stage for more in-depth studies of publication and collaboration behavior in the future that should give valuable insights into ways to increase scientific productivity in biomedical sciences.