Skip to menu | Skip to content |

Department of Information Science

CISR - Mission statement

The aim of the Centre for Interactive Systems Research is to continue work on the themes we have developed such as the Probabilistic Model of IR, and move into new areas such as Multimedia retrieval. Through our scholarship we will develop new methods to improve the searching experience of a diverse range of users, using real world user studies on different kinds of media. Our primary focus is on retrieval by text, of all kinds of information.


The History of the Centre

The Centre for Interactive Systems Research (CISR) is world renowned for its research and expertise in the area of information retrieval (IR). It has built up its profile since its inception, and has been particularly successful in the TREC series of conferences using the Okapi system.

The CISR was formed in 1987, with the aim of bringing together a group working on various aspects of information retrieval. Initial work was on theoretical approaches (particularly the probabilistic model) and on user-interface aspects. A major area of work at the time was OPACs for library catalogues.

A major step forward in 1989 was the arrival of the Okapi project from the University of Westminster. Okapi was designed as an OPAC, but incorporated some general text retrieval algorithms including some based on City work. It was used as the basis for a series of end-user experiments in relevance feedback, with two kinds of databases: library catalogues and scientific abstracts.

The second major step forward was in 1991-92, with the start of the TREC competition. Okapi was further developed as a flexible experimental engine, allowing both live-user experiments and laboratory-style batch tests. At the same time, further work was being done on the theory and algorithms, to make them more effective and robust with more heterogeneous data. The most prominent result of this work was the ranking algorithm now known throughout the international information retrieval community as Okapi BM25. The City team's success in TREC competitions from TREC-3 on has resulted in the wide adoption of BM25 by many research groups. At the same time, the team continued to work on user-interface aspects of retrieval.

More recent projects include TIPS, funded by the EU, and Collocation Linguistics/XML indexer for Okapi, both funded by Microsoft Research Cambridge.


The scope of Interactive Systems Research

The scope of the Centre focuses on access to, and retrieval of, information, which can be best expressed as a triangle:

IR triangle

The internal part of this triangle is the basic concepts and models of recorded information and its retrieval (for example the Robertson/Sparck Jones Probabilistic model and Cognitive models of IR). Each node of the triangle is related to each other, largely through the various theories which can be applied to IR. The information retrieval systems and algorithms 'node' reflects the majority of the centre's interests to date, the contexts & resources of information retrieval 'node' addresses either IR for particular groups (such as disabled people, dyslexics, engineers, scholars of the humanities etc) or within some environment (such as libraries, information centres) and finally the information user behaviour 'node' addresses the issue of how and why users do IR. There is a great deal of interaction between each of these 'nodes' and we give three examples below.

Much of the research done in the CISR has concentrated on textual data, particularly for ad hoc searching and information filtering tasks. We are taking our expertise in the user interface design and probabilistic models for information retrieval and applying them to new areas. For example we have expertise in the area of relevance feedback for text, and there is evidence that these methods can be used on other types of media such as Images.

The area of information retrieval for the disabilities is an important if poorly understood issue. A particular problem yet to be addressed in any detail is that of the print disabled, i.e. IR for blind or dyslexic people. We are currently working with the Centre for Human Computer Interaction Design on a project to understand the information seeking behaviour of dyslexic users.

Democratic or user based indexing is intended for use in a dynamic retrieval system which would allow users to contribute to the indexing and retrieval process. Our primary interests in the area of information media are to analyze and define meaning in relation to interpretative media and subsequently to design and develop retrieval systems that exploit 'levels of meaning' as generators of access points. We have submitted an EU proposal as part of a Semantic Web project to investigate some of these issues in a Ship Building application.

Centre researchers

Staff and associates

Research students

Key research themes

Resources