The Smart Media Institute (SMI) is a Research Centre established in the Department of Computer Science at UCD to carry out basic research into the next generation of Internet applications. This research involves using advanced techniques like machine learning, case-based reasoning and conceptual indexing to develop fundamental solutions to problems which will be faced by a wide range of future applications. This year, Enterprise Ireland has funded several new projects at the Institute.

One project, headed by Nick Kushmerick, uses machine learning techniques to perform information extraction, integration and retrieval. The Internet delivers mountains of information to the ordinary user’s desktop. In the future, users will require intellgent tools to sort, filter and re-present this information in a palatable form. For example, many commercial Internet sites display advertisements. Several systems have been developed for filtering out advertisements, but all require users to continually create and repair hand-crafted rules for detecting advertisements. In contrast, we have developed an AdEater system that uses machine learning techniques to automatically generate and maintain such rules. Users show AdEater example advertisements and the system generates rules that discriminate advertisements from regular images. Experiments demonstrate that AdEater’s learned rules are highly accurate, even when generated from relatively few examples.

There are several projects, involving Barry Smyth, Greg O’Hare and Mark Keane, that use case-based reasoning (CBR) in advanced Internet applications. CBR is a problem solving technique that relies on a corpus of past experience in the solution of future problems. New problems are solved by retrieving and adapting the solutions to similar problems that have been solved in the past and stored as cases in a case-base. CBR has been used in intelligent Web-site and e-commerce applications – Web sites that are capable of automatically learning about their users, and adapting their information content to match an individual’s preferences and needs. For instance, the PTV system ( http://ptv.ucd.ie ) is a personalised, web-based television listings service that learns about the viewing preferences of its users in order to deliver personalised electronic TV guides. The true potential of PTV becomes clear with the advent of Digital television – with hundreds of channels and thousands of programmes to choose from on any given day, the traditional television guide will be rendered useless, whereas PTV can deliver highly customised guides, listing only those programmes that a given individual is likely to watch. The technology underlying the PTV system can be easily adapted to personalise any Web site and, with this in mind, a new campus company, Changing Worlds Ltd ( http:// www.changingworlds.com ), has been established to bring this groundbreaking work to the market-place.

A third area, led by Joe Carthy, concerns the area of information retrieval, and deals with topic detection and tracking (TDT) in electronic texts. TDT techniques detect the occurrence of a new event such as a plane crash, a murder, a jury trial result, or a political scandal in a stream of news stories from multiple sources, and the tracking of a known event. This project aims to improve topic detection using conceptual indexing based on the WordNet computational lexicon, to identify the concepts underlying the terms that occur in texts, and use this for more advanced retrieval techniques than simple string matching.