Issues in Science and Technology Librarianship |
Fall 2009 | |||
DOI:10.5062/F40K26HX | ||||
Tips from the Experts
Author Identification Systems
Sciences Librarian
Science & Engineering Library
University at Buffalo
Buffalo, New York
abwagner@buffalo.edu
Abstract
Many efforts are currently underway to disambiguate author
names and assign unique identification numbers so that publications by a
given scholar can be reliably grouped together. This paper reviews a
number of operational and in-development services. Some systems like
ResearcherId.Com depend on self-registration and self-identification of a
researcher's articles. Some database producers are using a combination
of computer algorithms and manual intervention to assign author
identification numbers and thereby cluster publications as records are
entered into their systems. Searchers doing author name searches can
then use guided search features to review and select these clusters. Web
of Knowledge offers Author Finder while Scopus has a similar Author
Identifier feature. Various government agencies and non-profit
organization are also implementing or planning to implement additional
solutions such as CrossRef's ContributorID.
The Challenge at Hand
Author name disambiguation and the association of scholarly workswith the correct author have long been a problem for those wishing to
develop a comprehensive list of publications for individuals. Given all
the other advances in information retrieval and web technologies, some
scholars are wondering if we finally have the tools needed to tackle
this problem. Facets of this problem include:
- Inconsistent name formats caused by the authors themselves or editors
- Various transliteration systems, especially where different
non-Roman alphabet names result in the same transliterated Roman
alphabet name. - Legal name changes
- Cultural variants in the position of surnames
- Compound or hyphenated names
- The sheer volume of scholarly materials
- Highly similar names sometimes even doing similar work at the same institution.
- The large number of common names, especially certain surnames in many cultures.
number. This article discusses efforts underway to establish what would
be described by catalogers as an authority list of authors which can be
linked to a definitive set of papers published by each author. We will
focus on systems that allow authors to self-identify their own
publication list. In addition, we will briefly discuss some attempts by
database vendors to uniquely identify authors as records are processed
for inclusion in the database or during the formulation of a search
query.
Despite the power of computers and the ingenuity of these efforts,
the standard caveat of searching applies: Make no assumptions; try every
variant you can think of; and carefully screen the results.
ResearcherID.Com
In January 2008, Thomson Scientific, successor to the Institute forScientific Information, announced a completely free author
identification system called ResearcherID <http://www.researcherid.com/> (Thomson Reuters 2008). It is advertised as a global, multi-disciplinary scholarly research community.
During the free registration process, each scholar is assigned a
unique identification number and a permanent URL to one's personal
profile. You can then record the institutions you have worked for,
research areas of interest, descriptive text, keywords, role (e.g.,
academic researcher, student, or librarian), and contact information.
You have full control over what information appears in your public
profile. See, for example, this author's profile at <http://www.researcherid.com/rid/B-3784-2009>.
Most importantly, one can import one's publications list from Web of
Knowledge, EndNote/EndNote Web, or the generic RIS citation format
produced by many other personal citation managers. Though ResearcherID
can readily be used by researchers without Web of Knowledge (WOK)
access, there are certain advantages to being a WOK subscriber.
Citations imported from WOK have a link back to the database record and
you can view your citation metrics, including the h-index. This metrics
display is very similar to the Web of Science Citation Report feature.
Finally, a new tool will show authors which articles have already been
added to their publication list when they're searching WOK. One can
maintain up to three different publication lists per profile.
Anyone can search the registry and view public profiles to find
collaborators, review publication lists and explore how research is used
around the world. For example, searching for 'Wagner, A' leads to four
records, one of them being this author's profile.
At this point in time, there is no independent verification of
authorship for articles in ResearcherID. Though there have been no
reports of false claims, this may be a concern for some. On the other
hand, any vetted system would require large amounts of time and
resources which would mean it could not be free and potentially widely
adopted. This system is no different than so many other tools and
information on the Internet, requiring a mixture of individual
responsibility, an honor system, community policing, and for critical
work, verification by the searcher. Perhaps of greater concern would be
researchers that intentionally or unintentionally register multiple
times.
Resources at their web site include:
- A colorful fact sheet: http://isiwebofknowledge.com/media/pdf/ResearcherIDFS_web.pdf
- Web tools like an embedded research badge: http://isiwebofknowledge.com/researcherid/ridlabs/
- Online tutorials: http://scientific.thomsonreuters.com/training/rid/#recorded_training
AuthorClaim
AuthorClaim <http://authorclaim.org/>is an open source solution with the same objectives as ResearcherID, a
free author registration system that allows scholars to associate their
publications with a profile. The system is basically a clone of the
older RePEc Author Service <http://authors.repec.org/>
that has registered a majority of active economics researchers, over
20,000 at last report. Both services were created by the same person,
Thomas Krichel, who teaches at the Palmer School of Library and
Information Science at Long Island University.
During registration, one provides an e-mail address, full name, and
institutional affiliations. Name variations are generated by the system
which can be edited by the registrant.
The system then searches a set group of databases for candidate
articles. As of May 30, 2009, six databases are harvested: arXiv.org,
CiteSeerX@PSU, Current Index to Statistics, DBLP (computer science),
PubMed, and E-LIS (library & information science). In the case of
this author, three of my 10 refereed articles were found, not surprising
given the nature of the databases searched. Unfortunately, there
currently is no mechanism to manually add missing publications. This
likely would discourage many scholars from participating. On the plus
side, by using open access databases excepting one, Current Index to
Statistics, all publications are linked back to source database records.
Registration is completed and verified by an e-mail containing an
activation URL. Instead of an identification number, one receives a
permanent, public URL to one's profile. See, for example, the author's
profile <http://authorclaim.org/profile/pwa1/>.
The biggest drawback is that there is no way to browse or search
profiles, even by name. Unless researchers have publicized their
AuthorClaim URLs, you have no idea if they have registered. Since this
service is based on open source software <http://acis.openlib.org>,
Dr. Krichel hopes others will provide this capability. One could use
the ftp protocol to download the public profiles for all registrants,
but most researchers probably won't want to bother with that.
Vendor-specific Author Disambiguation Systems
MathSciNet, Web of Science, and Scopus all attempt to disambiguateauthor names, relying to a great degree on computer algorithms. Such
programs are time-consuming to develop and maintain, and the algorithms
are far from perfect. These efforts will be briefly discussed. They will
be mostly of interest to current subscribers of these services who can
directly test these features and come to their own conclusions about
their efficacy.
From its beginning in 1940, MathSciNet had an editorial policy to
uniquely identify and track authors. This started as a manual 3x5 card
operation. They now run a sophisticated program that automatically
identifies 80% of current authors with the other 20% requiring some
level of staff intervention (TePaske-King and Richert 2001).
Any searcher undoubtedly would desire that this level of authority
control was universally practiced by all database producers. However,
even within the narrow confines of mathematics, the process is very
expensive and reportedly requires a large number of staff.
Web of Science uses a combination of guided selections during the
search process and direct notifications from users regarding errors and
omissions. The challenge for any such approach is to minimize both Type 1
errors (including publications that are not by a given author) and Type
2 errors (missing publications that are in fact by a given author).
Like recall and precision, it is difficult to do both well at the same
time.
Web of Science's guided assistance is called "Author Finder" and
appears on the main search query screen. One enters a surname, first and
middle initials. One gets a list with a record count for various levels
of specificity such as Weinstein, B, Weinstein B*, Weinstein BA*, etc.
After choosing the desired set, the results are successively limited by a
pick list of broad subject areas, e.g., physical sciences, and then a
ranked list of institutions. Obviously, this feature is most effective
when one has researched the work history of the targeted author.
Scopus calls its feature Author Identifier <{http://www.info.sciverse.com/scopus/scopus-in-detail/tools/authoridentifier}>.
This web page links to a tutorial that demonstrates the capabilities.
As records are processed, computer algorithms identify authors and
assign them a unique number. This number is used to group together all
publications by a given author.
To search authors, one simply enters the last name and an initial or
first name. A list of the potential author names will be presented
along with name variants and number of publications. An Author Details
page gives additional information such as citing references and citation
metrics like the h-index. Search results can be further narrowed by
affiliation, city, country, and subject area. Enserink (2009)
reports that Scopus errs on the side of caution, providing multiple
clusters for what, upon further investigation, turn out to be the same
person. However, a searcher can temporarily group a number of separate
authors into a single entry.
In Development
Some discussions have begun about using OpenID as a researcher\publication identification system (Neylon 2009).OpenID is an open standard that allows users to create a single digital
identity and thereby log in to many different services automatically.
It has been suggested that OpenID information become a standard part of
article submissions or that the registration be expanded to include
publication lists in addition to the institutional affiliations already
captured. Though there is a degree of third party verification by virtue
of the requirement to register with an OpenID service provider, there
is no central party verification. As with ResearcherID, there still is
nothing to prevent a given person from having multiple OpenID's,
provided they have multiple e-mail addresses.
It will also be interesting to see how the UK Names Project
develops. They are currently running a pilot project designed to
provide UK institutional and subject repositories with a service that
will reliably and uniquely identify individuals and institutions. In
April 2009, they have received funding to expand the current prototype
into a more comprehensive pilot system.
CrossRef is developing a system called ContributorID. Since CrossRef
provides the infrastructure for persistent Digital Object Identifiers
(DOIs), ideally every article as it is accepted for publication would
not only be assigned a DOI, but each author would be assigned/report
their unique ContributorID. Hence, a permanent link would be established
between every paper and its authors. Very little information has been
released about this effort, though a brief news item appeared in CrossRef's quarterly newsletter.
It would be difficult to even list the universities, publishers,
non-profit organizations, and national agencies and governments
developing or implementing their own systems. Since 2007, Netherlands
has assigned each researcher a Digital Author Identifier. The National
Institutes of Health, specifically the National Center for Biotechnology
Information (NCBI), is working on a system so that it can link grants
and resulting publications.
Clearly over time, certain systems will prevail or merge. For now the
hope of a single worldwide researcher i.d. number is still unfulfilled,
but many projects are underway and are rapidly evolving.
References
Enserink, M. 2009. Scientific publishing: are you ready to become a number? Science 323 (5922):1662-1664.Neylon, C. (2009, January 20). A specialist OpenID service to provide unique researcher IDs? Message posted to: Science in the Open. http://blog.openwetware.org/scienceintheopen/2009/01/20/a-specialist-openid-service-to-provide-unique-researcher-ids/
TePaske-King, B. and N. Richert. 2001. The identification of authors in the Mathematical Reviews Database. Issues in Science and Technology Librarianship (31). [Online]. Available: http://www.istl.org/01-summer/databases.html [Accessed: June 25, 2009].
Thomson Reuters. 2008. Thomson
Scientific launches Researcher ID.com to associate a researcher with
their published works. [Online Press Release]. Available: http://scientific.thomson.com/press/2008/8429910/ [Accessed: June 25, 2009].
| Previous | Contents | Next |
Author Identification Systems
No comments:
Post a Comment