Title: Personalization in Digital Libraries

Introduction: The Internet nowadays is saturated with information that contends for people’s attention.  Information overload, a term coined by Alvin Toffler in his book Future Shock, refers to “the state of having too much information to make a decision or remained informed about a topic” (Toffler, 1970).  Over the years, organization of information has evolved from a rigid architecture of browsing and searching interfaces to a number of flexible innovations such as collaborative tagging, customization, and personalization (Morville & Rosenfeld, 2006).  This evolution contributed to the emergence of Web 2.0, an approach that technologists have adopted to facilitate online collaboration and user-generated content.  How does one take advantage of this explosion of “user-generated content”?  In the domain of digital libraries, how does one convert digitized information into knowledge, and eventually into wisdom?

Web 2.0 is about the empowerment of the people.  Semantic Web, which is playing a significant role in Web 3.0, focuses on content representation.  Berners-Lee et al. (2001) describes the Semantic Web as “bringing structure to the meaningful content of Web pages, creating an environment where software agents roaming from page to page can readily carry out sophisticated tasks for users”.  The potential of Semantic Web cannot be overstated.  Let’s illustrate an example in the healthcare industry.  Imagine a patient diagnosed with cancer at a late stage where conventional treatments are not possible.  His patient record is stored in an Electronic Medical Record (EMR) in which a standardized format has been established.  With Semantic Web, a software agent crawls through databases of clinical trials stored physically thousands of miles away and determines which of those are appropriate for the patient.  The software agent then provides the physician with a ranked list of recommendations based on the patient’s conditions.  The potential of Semantic Web is limitless.

Content representation is only half of the picture; user modeling is the other half.  A user profile is an abstract representation of an individual.  A typical user profile contains information that is both explicit and implicit.  Explicit information may include a person’s name, gender, age, race, and home address.  Implicit information, on the other hand, may include a person’s usage behavior, needs, and interests.  This information evolves over time based on a user’s interaction in an environment.  Personalization is a technique to map between content representation and user modeling.  When implemented properly, it is information management at its best.

In the digital library domain, the focus has been on information retrieval (i.e. accessibility, organization, and presentation), availability, and preservation of digital objects.  This project focuses on the enhancement of information retrieval (IR) in digital libraries using personalization.  Through stakeholder interviews, the potential benefits of personalization in digital libraries can be uncovered.  Stakeholders in a library system include managers, librarians, staff members, and end users.  The perception of each stakeholder groups on the value of personalization sheds light on how designers should proceed in implementing personalization features in digital libraries.  Secondly, affective measures (Nahl, 2005) have been proposed in a study to be used as quantitative performance indicators in an information-seeking environment.  In this project, we propose a usability study utilizing similar affective measures to quantitatively measure the perceived value of personalization in digital libraries.  Designers will in turn benefit from the results of such tests to learn what future improvements can be made to enhance IR in their respective systems.  Lastly, a new technique called attention profiling has emerged as a mechanism to systematically capture a person’s interest and tastes (Good, 2007).  We study the specifications of the Attention Profiling Markup Language (APML) to determine the feasibility of utilizing it for user modeling.

Downloads:

Report (1,924KB)

Presentation w/ Audio (28,876KB)

Presentation w/out Audio (2,294KB)

Poster (807KB)