Proposal Abstract

Digital libraries, which are curated collections of information objects of various kinds (e.g. books, journal articles, Web pages, images, and so on) in digital (that is, electronic) form, accessible over the Internet, are becoming increasingly important information resources throughout society. A major problem in supporting effective use of digital libraries for users of all sorts, is that digital libraries have typically been designed to provide a uniform style of access for all users. However, it has been shown that various characteristics of users of digital libraries (and other information systems), and their contexts, such as cognitive ability, experience with the system, knowledge of the topic of a search, goals which have led to their information seeking behaviors, task in which they are engaged, all affect how searchers interact with a system, and what would constitute an effective response from the digital library for any particular user at any particular time. The proposed project addresses the issue of making interaction with digital libraries more effective and engaging through personalization of support for interaction to the individual searcher, and that searcher's context.

There has been some substantial research, and practical experience, in aspects of personalization of interaction with information systems, such as book or movie recommender systems, based on a variety of criteria. However, there has to date been little research done on how to unobtrusively discover relevant characteristics of a searcher and the searcher's context, and integrate the knowledge thus gained in determining how best to tailor any particular search experience precisely to that searcher's current situation. In addition, most personalization research has been carried out by large companies, such as Google or Amazon, with the intention of doing the personalization at the site of the information system. This approach has obvious privacy problems.

Our proposal is to develop a personalization assistant, a software application which will reside on the searcher's own computer(s), monitor the searcher's past and current behaviors, and tailor the entire searching experience with any given digital library or other information system, at any given time, according to the knowledge of various characteristics of the searcher, the searcher's context, and the specific situation at that time. The proposed project begins with analysis of previously collected data on people's searching behaviors in relation to their judgments of usefulness of digital information objects; continues with a series of experiments which study the problems of unobtrusive collection of relevant information about searchers, the interpretation of that evidence in an integrated fashion, and the use of that evidence for effective personalization; and concludes with the construction and evaluation of a prototype personalization assistant for personalized interaction with digital libraries. The project is scheduled to begin in October 2007, and to be completed in September 2010. At the conclusion of the project, the prototype personalization system will be made available to the digital library and information retrieval research communities, and to digital libraries in general, for further research and evaluation, and for implementation in operational digital library settings. Overall, if results of the project follow our expectations, it should result in a quite different model of how people can interact with digital libraries and other information systems, which we believe will profoundly enhance the use of such systems. In particular, we expect the personalization assistant to lead to a quantum jump in the effectiveness of access to information in Digital Libraries.

Full Proposal

The PooDLE Project is funded by the Institute of Museum and Library Services.