In Research

• Introduction
• Theory
• Vision and Interaction
  • Vision
  • Interaction, Media and Communication
• Information Engineering
  • Risk Assessment
  • Information Retrieval
• Public Engagement
• Interdisciplinary research
• HUB research magazine
• Technical Reports
• Augmented Human Interaction Lab
• Contact us

See also:

• Women in Computing Research
• PhD degrees
• People
• News and events
• Business
• Facilities/resources

Information Engineering

Introduction

The Information Engineering group specialises in Risk Assessment (led by Professor Norman Fenton) and Information Retrieval (led by Professor Mounia Lalmas). The group's risk assessment work is unique in the UK in its focus on problems of decision-making under uncertainty, and is world-leading in the development of Bayesian net technologies in collaboration with the spinout company Agena. The impact of this work has been enormous: over 3,000 practitioners and researchers worldwide use software and models developed by the group in applications as diverse as defence, banking, and transportation. The work has also been used in a patented system for TV program recommendation, and in major legal cases.

The Information Retrieval work is widely recognised as among the best in Europe. The results from its broad portfolio of research are now being applied in the context of distance learning, digital libraries, interactive TV, and business information domains. The research led to the development of HySpirit, a technology for building search tools, and to a spin-out, Apriorie Ltd. The group runs the international evaluation initiative for content-oriented XML retrieval (INEX).

International quality researchers in this group include: Professor Norman Fenton, Professor Mounia Lalmas, Dr William Marsh, Dr Christof Monz, Professor Martin Neil, Jane Reid, Dr Thoman Roelleke and Dr Tassos Tombros. Highlights of our research include:

Software Risk: Professor Norman Fenton, Dr William Marsh and Professor Martin Neil have produced some revolutionary extensions to their earlier highly-cited work on software defect prediction using Bayesian Nets. In particular, their work has led to radical improvements in the accuracy of software defect prediction and has led to the ability to apply such predictions in arbitrary software environments. Numerous companies including Philips, Motorola, QinetiQ and Israel Aircraft Industries have used the models and tools resulting from this work. For example, Philips reported 95% accuracy in software defect prediction, giving them greater confidence in decisions for testing and release of components in the critical area of embedded systems in consumer electronics.

Bayesian Network Modelling: The team have also made several radical advances in Bayesian network modeling, solving some classic problems. Their approach to simulation means that Bayesian networks can now produce accurate results with continuous variables (partly solving the so-called dynamic discretisation problem). Their research on modelling ranked nodes solves the difficult problem of intractability of building realistic models based on expert judgements; Their work also combines multi-criteria decision making with Bayesian uncertainty reasoning to solve complex problems involving causally dependent preferences. Many of these advances and algorithms have been implemented in the commercial software tools Hugin and AgenaRisk, resulting not only in more accurate models, but also the possibility of modelling problems that were previously beyond the range of Bayesian analysis.

Risk assessment: Our pioneering work in risk assessment using Bayesian networks has had applications outside software. In particular, it has produced novel solutions to problems in vehicle reliability assessment. This work led to the revolutionary TRACS system that is used routinely by QinetiQ for evaluating vehicle tender bids on behalf of the MOD, railway safety and operational risk.

XML retrieval: Many document classes are now based on XML and the techniques evolved for ordinary text and hypertext have problems in both retrieval and evaluation. Professor Mounia Lalmas has led an international effort, INEX, to solve these problems. Her extensive investigations, both empirical and theoretical, led to the development of an evaluation methodology for XML retrieval and new metrics for measuring retrieval effectiveness. With these, more than 60 organizations worldwide were able to advance the state-of-the-art XML retrieval research.

Probabilistic retrieval models: Dr Thomas Roelleke's research examines the foundations of retrieval models, leading to a new view on probabilities, namely to distinguish between occurrence and informativeness probabilities. This view triggered a reformulation of the most established probabilistic retrieval modeland has influenced the design of future IR systems. The formalism he proposed has now been adopted by the leading information retrieval group, CWI Amsterdam, to implement complex retrieval tasks.

Document summarisation and cluster-based retrieval: Dr Tassos Tombros has provided significant evidence that document content alone is not sufficient for summarisation, and relevance analysis and the incorporation of structural features is also crucial. Building on his BCS/CPHC Distinguished Dissertation'ss winning thesis, he has shown that highly effective cluster-based document retrieval can be achieved by selectively considering inter-document relationships, and given a new model for measuring these relationships. This work contradicted previous orthodoxy, which disregarded cluster-based retrieval, and the approach has been widely adopted, for example at the world leading group at the University of Massachusetts.

Cross and multi-lingual retrieval: Dr Christof Monz's work on the evaluation of approaches to retrieving from multi-lingual knowledge bases has been widely influential. His work advances the state of the art in providing uniform platforms to work across language groups in support for queries from those who are not language experts.

Top

Industrial collaboration

The group have very strong industrial collaborations. Highlights include:

• Professor Norman Fenton led the EPSRC SIMP project (Systems Integration for Major Projects), whose key industrial partner was BAe Systems, for whom Professor Fenton and his team developed a model for assessing risk in one of BAe's most critical defence projects.
• Dr Martin Neil led the EPSRC SCORE project (Sensing Changes in Operational Risk Exposure) whose industrial partners were NATS, ERA, CAA, and Agena and is leading the DyFusion project whose industrial partners are Agena, QinetiQ and Motorola.
• Much of the Bayesian network research has been brought to market by the company Agena, in which Professor Norman Fenton and Dr Martin Neil are directors. They have been involved in the continual development of the AgenaRisk software system that can be used for a wide range of risk assessment problems. The tool has hundreds of commercial users world-wide including: Royal Bank of Canada, QinetiQ, Motorola, Siemens, Orange, Philips, Tellabs, General Dynamics, Israel Aircraft Industries, TNO (Holland).
• In addition over 3000 research users in 117 different countries have taken advantage of free access of the various Bayesian net models and software developed by the Group.
• Professor Norman Fenton and Dr Martin Neil have also engaged in a range of risk assessment consultancy projects for companies including Royal Bank of Canada, ERA, Motorola, Philips, QinetiQ; National Air Traffic Services; UK MoD DSTL and Railtrack. Marsh provides regular training to ERA Technology Ltd and other organizations on state-of-the-art systems and software safety.
• Professor Norman Fenton is an expert witness in risk assessment including a civil case in 2006 involving safety critical software intended for the London Underground Tube system, and a criminal case in 2007/8 involving multiple murders.
• Dr Thomas Roelleke's research led to the development of HySpirit, a technology for building search tools, and to a spinout, Apriorie Ltd, used at information management projects at the BBC and Credit-Suisse and at Fraunhofer Darmstadt and University of Duisburg.
• The INEX project that Professor Mounia Lalmas leads has over 60 participating organizations worldwide. Part-funded by DELOS (an EU network of excellence in Digital Libraries), it is allowing XML retrieval research to expand by providing means for the entire XML retrieval community to evaluate numerous aspects of XML information access, ranging from system performance to user interaction. Most published work and PhD theses on XML retrieval makes use of INEX to validate their approaches.
• Professor Mounia Lalmas is funded through the Yahoo! Research Alliance Programme, and Dr Christof Monz's work on cross language retrieval is funded by the UK Government.

Top

Academic collaboration

• In EPSRC projects SCULLY, SIMP, and SCORE the Group's academic collaborators were the Universities of Surrey, City, Manchester, and Liverpool.
• The new interdisciplinary EPSRC project, Renaissance, of Professor Mounia Lalmas is in collaboration with the University of Glasgow and the Open University.
• Through INEX, the group has strong collaboration with all major international IR groups.
• The group (together with the Department of Electronic Engineering) is a founding member of MMKM, an EPSRC network on Multimedia Knowledge Management, which includes Imperial College London, and six other UK universities.
• The group works with the Queen Mary Medical School on Bayesian inference, and Queen Mary Electronic Engineering on information retrieval from TV broadcasts.
• Funded visiting affiliations overseas have included Professor Norman Fenton (Haifa, Affiliate Professor), Professor Mounia Lalmas (Paris 6, Barcelona), Dr William Marsh (COMSATS, Pakistan), Dr Christof Monz (CMU), Dr Thomas Roelleke (Paris 6), and Dr Tassos Tombros (Patras).

Top