Mohammad Ghoniem’s website

© Mohammad Ghoniem – février 2007

Research Interests

Saturday 27 January 2007, by Mohammad Ghoniem

All the versions of this article:

In the last six years, I have gained a strong expertise in the field of Information Visualization. This discipline provides compact computer-supported visual representations of abstract multi-dimensional data along with appropriate interaction techniques and user interface in order to help explore, reason, communicate and eventually make decisions about massive volumes of complex data. Depending on the characteristics of the data at hand, whether structured (hierarchies and networks), sequential, quantitative, nominal, ordinal etc., different visualizations and human computer interaction techniques provide invaluable exploration tools and understanding aids. I have acquired this experience through my MS and Ph.D. theses both in the industry at ILOG France and in French academic research labs.

On a personal ground, I have gained expertise with regard to the information visualization applications and challenges, as well as a practical knowledge of the problems and requirements met while designing such demanding computer software. My work can be mobilized fruitfully in all I.T. demanding industries where large volumes of complex data are produced and stored in view of further analysis and decision making.

Within the French National Project “OADymPPaC”, my skills have been applied in particular to the monitoring of large dynamic networks produced by constraint-based solvers such as ILOG Solver, Cosytec Chip++ and GNU-Prolog. I designed a novel matrix-based visualization tool that succeeds very remarkably in the visualization and monitoring of large dense co-activity networks, whereas traditional node-link representations fail. Thanks to information visualization tools, constraint-oriented programmers and experts have now gained invaluable insight with regard to the intimate behavior of solvers and are able to debug, analyze and tune their algorithms.

As a post-doctoral researcher at the Charlotte Visualization Center, University of North Carolina, Charlotte (UNCC), my research belongs to the current research agenda of the South-East Regional Visualization and Analytics Center (SRVAC). I am particularly interested in information visualization and human-computer interaction techniques well-suited for the interactive exploration of large datasets. My current work focuses on the following topics:

Semantic browsing of multimedia data

Recently, the number of TV broadcasters has dramatically increased and, at the same time, feature rich multimedia contents have been made available for a very large audience who can also access to cheap large storage devices. This trend has brought about the need to explore and make use of large repositories of video data, whether private or publicly available, for various types of application.

Therefore, I am currently working on the visualization of large repositories of video data, such as TV news reports, augmented with closed captioning and meta-data generated by automated video content segmentation programs. These data raise many interesting questions such as the monitoring of topic changes over time in the media, the detection of emergent stories, the detection of outliers, as well as the comparison between various broadcasters in the way they treat information.

An immediate application of this work may be the automated compilationof a visual and interative press review from various multimedia sources, or the visualization and classification of a private video collection.

Financial data visualization

The computerization of financial transactions taking place in market places, financial institutions, energy markets etc. raises the need of making sense of large volumes of transactional data involving various types of actors. At stake is the capacity of analysts to read the evolution of markets across time, to establish a clear picture of given industry according to various indicators or points of view, or even to detect and fight against fraudulent transactions. These requirements could be dictated by competitivity and benefit goals, or by risk managers who are required by law to report dubious activities going through their financial network lest their institution would be closed or severely sanctioned.

In this context, I have been involved in a project involving a major American bank aiming at the design of information visualization tools for wire transaction data capable of scaling up to long periods of time at the rate of several hundred thousands of transaction per day.

The visualization of large time-varying networks

For my Ph.D. dissertation, I worked on the visualization of large co-activity graphs in support of the analysis and debugging of constraint-oriented programs. I exhibited the effectiveness of matrix-based representations of graphs compared to more traditional node—link diagrams, when dealing with large and/or dense networks.

This technique has also been successfully applied in other application domains such as social networks, and telecommunication and computer networks. Such data structures raise challenging problems with regard to scalability, readability, et become very difficult when time-varying attributes are attached to the network’s vertices or links.

A more detailed description of this problem can be found in my publications between 2001 and 2005.

The evaluation of information visualization techniques

The proposition of matrix-based representations of graphs as an alternative to node—link diagrams led me to set up a controlled experiment in order to compare the readability of both representations. Such initiative includes the choice of tasks to be accomplished by the subjects, the precautions in order with regard to the equivalence of these representations interaction wise, the characteristics of the dataset used in the experiment, and levelling undesirable effects such as memorization or weariness. Eventually, comes the statistical analysis of the collected results.

For the sake of the aforementioned evaluation, I implemented a generic evaluation framework for visualization techniques, that handles the view generation, the question display, the collection of answers and their timing.

Knowledge visualization

The design of visualization systems involves a fair amount of knowledge related to the application domain where the data come from, as well as information visualization art and techniques. Modeling such knowledge according to state of the art techniques in the domain of KDD offers interesting perspectives since it becomes possible to embed knowledge and inferecing capabilities in view of building more "intelligent" visual analytics tools. For example, I have worked on modeling the geopolitical knowledge that is often involved in human activity (e.g. economy and culture) and am currently investigating its use in various applications such as the monitoring of bank transactions and the analysis of TV news reports.

RSS 2.0 [?]

Espace privé

Site réalisé avec SPIP
Squelettes GPL Lebanon 1.9