Research Interests

This page describes the research activates undertaken over the past few years, the related publications can be found under Richard Freeman’s publications section

Richard’s moto is “Making use of open source software and public datasets to accelerate the development of research prototypes”

2010 – 2011 : Global Recruitment Systems

Richard is currently working as a lead global architect delivering search, multi-channel communication and advertising projects for a IT transformation programme across 32 countries. He is focusing on building a replacement recruitment system based on Enterprise Search, Enterprise Service Bus, Enterprise Content Management, and Business Process Management suites, and running different Proof of concepts for a global recruitment firm. Richard is also looking at designing a highly available system that organises web content into a automatically derived taxonomy. This is based on his earlier PhD work using Self-Organizing Maps.

2009 : Neural-based Web Feed Aggregator

Richard focused his efforts on enhancing web feed aggregators. With the rapid and dramatic increase in web feeds published by different publishers, providers or websites via Really Simple Syndication (RSS) and Atom, users cannot be expected to scan, select and consume all the content manually. This is leading to an information overload for consumers as the amount of content increases. We apply the topological tree method, to dynamically identify categories, on financial and business news feed dataset. The topological tree method is used to automatically organise an aggregation of the financial news feeds into self-discovered topics and allows a drill down into sub-topics. The news feeds, organised using the topological tree method, are discussed against those of typical web aggregators. A discussion is made on the criterions of representing news feeds, and the advantages of presenting underlying topics and providing a clear view of the connections between news topics.

2008 – 2009 : Applying research and complex event processing to insider dealing detection

Richard worked on an evolving intelligence-based market abuse detection system with real time price alerting, for financial instruments and in line with UK and EU legislation. Prior to this he helped a client formulate a strategy for creating a global group wide Centre of Excellence capability for shared services and delivery. In addition he worked on a large regulation and compliance piece of consulting work. At the same time he has been publishing a number of white papers including some relating to enterprise search, BPM, ECM and participating in a number of IBM events.

2006 – 2007 : Growing industry focused and Proof of Concepts

Richard worked on a number of ECM and BPM solutions for large financial services companies. He helped implement a novel paperless claims handling system around an advanced document management system, and was leading an offshore team in the delivery of a prototype aiming to explore high risk architecture components. In additon Richard has pubished a number of papers at international conferences. He also took an active role in the IBM community by attending and networking at the Innovation Series and Partner Events, as well as running IBM FileNet sessions with some clients and collaborating closely with the IBM vendor on RFPs and proof of concepts.

2004 – 2005 : Part time research and client focused delivery

In 2004, Richard joined one of the leading multinational IT services and consulting companies as an IT consultant in Information Management and Enterprise Content Management. In his spare time he has been continuing his own research related to his PhD topic. He has been publishing and presenting articles at major international conferences. For this, he has been employing best software practices to develop a web application that can query a web search engine and crawl the resulting pages. The documents are then organised using an improved online version of the adaptive topological tree method.

2004 : PhD write-up, enhancements to Topological Trees and further experiments/results.

In 2004, Richard entered the most difficult part of the Ph.D., the final write-up stage! He had a lots of ideas but no enough time to realise them all. He will publish these in future papers once his thesis is complete, as he will first write-up what he has done up to now. Some of the enhancements will be used for efficient information access, knowledge discovery and decision support.

He has also proposed the topological tree a new kind of growing tree representation with features from graph analysis and generated through his proposed method. These adaptive topological trees use the hidden features present in all tree representations. It has novel properties making it well suited for exploring large disorganised datasets, discovering complex hidden relations and searching large information spaces of document sets.

The thesis details how our work on the Adaptive Topological Tree Structure relates to the areas of enterprise information portals and business intelligence. The methods generates a new type of tree view, which resembles a graph structure yet can be efficiently and naturally represented as a tree and using XML.

2003 : Advanced prototypes in Java and journal publication of Topological Tree.

The entire application was ported to Java for flexibility and reuse. The implementation has gone further than the Visual C++ version and now includes other SOM implementations such as the basic SOM, Growing Grid and Growing Hierarchical SOM to allow a comparison between methods. The taxonomy generated by the proposed method has proven to be robust, scalable and extremely useful for browsing & performing knowledge discovery on the unexplored collection of documents. He also started to take a more focused look at using my algorithm in enterprise content management and knowledge management systems.

2002 : Initial prototypes in C++ and conference publications.

He is now focusing on document organisation, analysis and retrieval using Self-Organising Maps, which may also be written as Self-Organizing Maps or simply SOM. The SOM is used to perform document clustering where, textual documents are automatically sorted into a dynamically built hierarchy of topics, solely using the documents contents. He is currently working on large sets of web documents, created by book publishers, and which describe their books. A fully functional Windows OS application was implemented in Visual C++ MFC, to automatically cluster these documents into a hierarchy of topics or taxonomy. The method we call Treeview SOM (TV-SOM) allows documents to be clustered very rapidly, using a minimum number of nodes, into a treeview hierarchy of topics and without the use of any further post-processing (unlike other existing SOM methods). This type of view is optimised for user navigation and understanding.

This work is closely related to knowledge management, information retrieval and text mining. To build an effective system, he believes, that a true multi-disciplinary investigation has to be undertaken in the three areas of:

* Linguistics: such as semantics, discourse analysis, natural language understanding.
* Statistics: clustering, compression and co-occurrence analysis.
* Computing techniques: neural networks, Indexing and Information Retrieval.

2001 : In depth review of information retrieval and document clustering. Detailed planning and layout of ideas.

The original MPhil title was:

“Neural Networks for multimedia data exploration and visualization”

Technology for improving computing power has focused much on increasing processing speed. Whilst increasing computing efficiency has also become a demanding issue in today and future computing world as resources and information volume grow as astonishing speeds both locally and across the internet. Efficient data management and intelligent information retrieval methods will stand as important functions in future computing and information systems. The project will address and seek the novel methods in exploring and demonstrating multidimensional, multisensory and multimedia data by means of neural networks, self-organizing systems in particular. As the human cognition system, which often has multisensory inputs, has various levels of pattern association and effective information retrieval. It also is error resilient. The aims and objectives of the project are to achieve a better understanding of the human cognition system in relation to pattern association, to model the process using neural networks, fuzzy logic and genetic algorithms, and to produce corresponding novel computer systems for managing web or multimedia data/ documents. The project will also involve studies of statistical methods such as multidimensional scaling and factor analysis and development of computer software.