Computer Science Department
School of Computer Science, Carnegie Mellon University
Finding lists of People on the Web
Also appears as
Keywords: Information retrieval, data linkage, data mining, privacy,
Institute for Software Research International
Technical Report CMU-ISRI-03-104
Among the vast amounts of personal information published on the World
Wide Web ( Web ) and indexed by search engines are lists of names
of people. Examples include employees at companies, students enrolled
in universities, officers in the military, law enforcement personnel,
members of social organizations, and lists of acquaintances.
Knowing who works where, attends what, or affiliates with whom
provides strategic knowledge to competitors, marketers, and government
surveillance efforts. However, finding online rosters of people does
not lend itself to keyword lookup on search engines because the
keywords tend to be common expressions such as employees or students.
A typical search often retrieves hundreds of Web pages requiring many
hours of human inspection to locate a page containing a list of names.
As a result, people may falsely believe online rosters provide more
privacy than they do. This paper presents RosterFinder, a set of
simple algorithms for locating Web pages that consist predominately
of a list of names. The specific names are not known beforehand.
RosterFinder works by identifying rosters from candidate Web pages
based on the ratio of distinct known names to distinct words appearing
in the page. Accurate classification by RosterFinder depends on the
set of names used. Results are reported on real Web pages using:
(1) dictionary lookup employing a limited set of known names; and,
(2) dictionary lookup on utilizing an extensive set of known names.
Privacy implications are discussed using the example of FERPA and
online student rosters.
*Institute for Software Research International, School of Computer
Science, Carnegie Mellon University