Luz M. Quiroga
University of Hawaii
Reducing information overload is the main goal of Information Filtering (IF) and it has been recognized as one of the priorities in the development of current web-based information systems. IF systems are meant to deliver personalized information, acting as personal information agents that recommend relevant (filtered) documents based on their clients’ information preferences and needs (profiles).
Recommendation technology is being presented as a new paradigm of search where relevant items find the user instead of the user explicitly searching for them (http://recsys.acm.org/2009). New trends in Information technologies such as social networking and mobile devices are making personalization research and practice a priority.
Libraries have been offering personalized system in services such as: selective dissemination of information, alerting services for a long time. Customer and marketing research has also a long tradition. With the advances in information technology, personalization has evolved, now covering more sophisticated ways. Collaborative filtering, recommender systems, personalized help systems, social filtering, social data-mining systems, and user-adaptive systems can be collectively called information-filtering (IF) systems. Today, personalization is everywhere, in every industry and service, from marketing to health, travel, education, entertainment, etc.
IF researchers contend that a conceptual framework for the design of IF systems comes from two well established lines of research: Information Retrieval (IR) and User modeling (UM). The course covers theories, research and current practices in these two fields, including modeling and representation of documents, queries, user preferences, and user-system interaction.
The first part of the course includes IR models for searching: set theoretic models (e.g. Boolean model) and algebraic models (e.g. vector model). Emphasis will be given to query languages and protocols as well as to relevance feedback and strategies for query expansion and reformulation using, for example, different types of thesauri, metadata and markup languages (SGML, HTML and XML) that provide information on the document structure, format and semantics will also be included as part of the study of Web Based Information Retrieval and Filtering. Students will learn about system and user based retrieval performance evaluation and will experiment with benchmark tasks and reference test collections.
The second part of the course will mainly focus on user modeling. Although IF could be considered an application of IR, there is a major distinction: the existence of a highly individualized profile that is a representation of relatively stable user information preferences and needs. Profiles can be considered as user models and will be the center of this second part of the course which will review core topics in IF research including user modeling in IR and IF systems, acquisition of user profiles, personal ontologies, IF taxonomies, IF performance evaluation and Personal Information Management (PIM).
Baeza, R., Ribeiro-Neto, B. 2011. Modern Information Retrieval, 2nd ed.
Link to Syllabus: