On co-authorship for author disambiguation

Abstract

Author name disambiguation deals with clustering the same-name authors into different individuals. To attack the problem, many studies have employed a variety of disambiguation features such as coauthors, titles of papers/publications, topics of articles, emails/affiliations, etc. Among these, co-authorship is the most easily accessible and influential, since inter-person acquaintances represented by co-authorship could discriminate the identities of authors more clearly than other features. This study attempts to explore the net effects of co-authorship on author clustering in bibliographic data. First, to handle the shortage of explicit coauthors listed in known citations, a web-assisted technique of acquiring implicit coauthors of the target author to be disambiguated is proposed. Then, a coauthor disambiguation hypothesis that the identity of an author can be determined by his/her coauthors is examined and confirmed through a variety of author disambiguation experiments.

Introduction

As a tremendous amount of person-related information is available on the web or other electronic media, people want to find persons of interest frequently using the names as queries (Guha & Garg, 2004). However, there exists many-to-many mapping relationships between persons and their names. A person may have multiple names, and different persons may share the same name. For example, a person named David S. Johnson can write his name as David S. Johnson, David Johnson, D. S. Johnson, or D. Johnson, etc., and there may exist two or more different persons named David S. Johnson. When a single person is viewed as an object having a unique meaning, different names signifying the same person can be considered as synonyms, and the same names indicating different persons homonyms.¹

The above many-to-many mapping between persons and their names may severely deteriorate the effectiveness of the person search. When a person name is submitted to a web people search system (Wan, Gao, Li, & Ding, 2005) that finds persons in web pages, the web pages that have synonymous names of the input query may not be retrieved, lowering the recall of the person search, and the retrieved pages may contain numerous homonymous names of the input query, decreasing the precision of the search system. Moreover, synonymous and/or homonymous person names may hinder the correct representation of person-related social networks such as coauthor and citation networks.

Thus, identifying synonymous person names and resolving homonymous ones are crucial to any person-related search or representation systems. Thus far, much research has been conducted to attack the identification of synonymous person names under the terms of record linkage, merge/purge, duplicate detection and elimination (Elmagarmid et al., 2007, Gu et al., 2003, Winkler, 2006). Matching of synonymous names employs similarity measures which mainly rely on name-internal features such as overlapping characters or tokens (first, middle, last name) of the two names to be compared. On the other hand, disambiguation of homonymous names has relatively recently received a growing attention with the advent of the semantic web and social networks. It is highly dependent on name-external features: domain-independent biographic features (Guha and Garg, 2004, Mann and Yarowsky, 2003, Wan et al., 2005) such as birth data/place and e-mail/postal addresses, and/or domain-specific contextual features such as coauthors and paper titles in case of bibliographic data.

This study addresses the resolution of homonymous author names appearing in citation data. As disambiguation features, previous works have employed coauthors, titles of articles, titles of publications, and years of publications that constitute basic citation data. Titles of articles may epitomize research areas of their authors, thus under the assumption that namesakes do not heavily share their research areas, title similarity between articles may be employed in resolving homonymous author names. The same also applies to the case of titles of books. In addition to the above citation-internal features, some have utilized citation-external features such as abstracts, self-citations, and citation URLs. When the full text of articles is available, additional features such as e-mail addresses, affiliation, and keywords can be extracted and applied to the author name disambiguation.

We believe that of the above features, co-authorship is the most reliable and decisive from the viewpoint of discriminating the identities of authors, since it implies real-world acquaintances among authors. For instance, when one citation contains D. S. Johnson and C. J. Date as its authors, we can say that a D. S. Johnson in the citation indicates the D. S. Johnson whom C. J. Date knows. When another citation has also D. S. Johnson and C. J. Date, we can further state that “D. S. Johnson”s in the two citations are the same individual, namely, the D. S. Johnson whom C. J. Date knows, under the assumption that “C. J. Date”s in the two citations are not different persons. Other citation features are not as much person-related as co-authorship.

This study concentrates on investigating the sole effect of co-authorship information on the resolution of homonymous author names in bibliographic data. In particular, we start with a hypothesis stating that the identity of an author is characterized by his/her coauthors. Next, a web-based technique of gathering coauthors is proposed to supplement implicit coauthors not found in known citations. Then, our hypothesis is investigated using large-scale test data from the following viewpoints:

• Is it helpful to enrich coauthors with web-assisted collaborators in disambiguating namesakes?

• Is a person identified more accurately by more coauthors?

• How probable is it that co-authorship succeeds in disambiguating authors?

• Is it beneficial to use coauthors of coauthors in resolving authors?

The scope of co-authorship that this study handles is limited to academic co-authorship. In addition, the evaluation of co-authorship is performed exclusively using a test set in Korean. The remainder of this article is organized as follows. Section 2 summarizes related works. Section 3 describes the usage of co-authorship for author name disambiguation. Next, a web-assisted method of expanding coauthors is proposed in Section 4. Then, evaluations and concluding remarks are given in Sections 4 Web-based acquisition of coauthors, 5 Evaluation setup.