IRSIT is one of the 8 NRIs reporting to the SERST. It was set up in 1987 and, under the direction of Dr. Karima Bounemra Ben Soltane, it gathers about 55 full time researchers with an annual budget of the order of 1 650 kECU. It performs R&D in three major areas :
(1) Telecommunications and Networks
(2) Decision Support Systems
(3) Arabisation and Man-Machine Communication
The basic data related to IRSIT activities are given in Annex IV. They also suggest the general comments underneath:
(1) Telecommunications and Networks
(a) Interconnection of networks : the team have been involved in the interconnection of the "Reseau National de la Recherche et la Technologie" (RNRT) with the agriculture and the PTT ISDN networks and externally to Internet. They have monitoring and performance assessment skills.
(b) Telematics: this area is concerned with videotex systems. ALFICOM is a videotex presentation system and has related systems ARRASSEM for page composition / editing and the system is used in project T.A.P. for videotex access to the databases (in Arabic) of Tunisie Agence Presse. The work is clearly motivated by the PTT and has been done competently.
(c) Multimedia :
(i) Virtual Aquarium: a joint project with INSTM (Institut National des Sciences et des Technologies de la Mer) which provides easy-to use educative multimedia presentation of the information of INSTM. This project has an excellent user interface for education or information and presents the information in a well-structured and lively way. It is comparable with EU projects.
(ii) Audio-visual examination of the Highway Code: a project using structured multimedia with audio questions to collect multiple-choice answers and provide a scored proforma. Death and injury from Road Traffic Accidents is costly and this project has demonstrated a route for reduction. The techniques are similar to those in EU countries although there is no educative (feedback using AI) function - however, this was not specifically required.
(iii) OCSAD: a Medical Diagnostic Support System: this is a well-engineered system with multimedia (multiple windows, scroll-boxes, buttons and graphical representations) based on two levels of knowledge: medical knowledge with scored relationship strengths (Bayesian network style) and individual patient information. The system allows controlled and mediated interaction between these two and guides the clinician in diagnosis (stating why certain diagnoses do not fit the medical knowledge, for example, and indicating where additional information would help). The project seems to ignore much previous EU work in the field and is built from scratch with home-built software - because the cost of packages (e.g. expert system shells) would preclude its widespread deployment throughout Tunisia.
(d) Space Technology : this area is just taking off. The team have organised conferences and exchange / collaboration relationships with CNES (France) and others. The team participates in project COPINE - Eurafrican satellite communication. This is clearly an area for future development with both telecommunications and multimedia (image processing) aspects.
(2) Decision Support Systems (Salah Benabdallah)
(a) GIS and resource management decision support systems: there are several projects sharing the utilisation of the ARCINFO GIS software but extending it in different dimensions.
One project concerns socio-economic thematic mapping of Tunisia with point and click access generating a window with a database table displayed with information relating to the area or point (region or village); the information concerns sizes and costs (e.g. of healthcare or education). The database has annual versions so time-trends can be detected. A similar project concerns the Golfe de Gabes and its ecology, especially its fish stocks. Marine seabed mapping (using sidescan sonar among other techniques) is providing detailed information used by the Fisheries ministry and the Environment ministry. These two projects have relatively little research but are clearly productive and useful. Detailed printed cartographic maps are produced on paper using their own CAP (Computer Aided Printing)system.
A large project (Majed Khalfallah) concerns decision support with GIS in telephone distribution networks. The GIS gives the visualisation but the major resource is the database which is manipulated and navigated using genetic algorithms to find optimal routing of pipes or cables. The genetic algorithms are developed in the INCO project ISC-MED-35 EGOIST (2 years 150 kECU). Project GOLD (Genetic algorithms for Optimisation of Looped water Distribution networks) uses the same techniques. Project Mathusaleh concerns raster to vector transformation including tidying of the generated vectors. The 3 major projects are novel and well-performed research applying and developing the latest IT techniques to real-world problems.
(b) Decision Support for industry: The team demonstrated and described GRAM, a system for optimisation of preventative maintenance on the Tunis trams of one of the co-operating partners. The demonstration had some technical problems and there was no clear explanation of the underlying operational research technology - it appears to be linear programming rather than the more modern constraint logic programming. Nonetheless, the system appears to be satisfactory but (except for the modern Windows interface) reminds of Operational Research programs from the seventies. It is not clear exactly which technique is being used - there is a possible link with genetic algorithms in the GIS part of the team.
The team also presented Equinium: a production optimisation system for the textile industry. The system optimises production by decomposing overall tasks, allocation to workstations (person plus machine) and rescheduling for optimal throughput. The algorithms were not described in detail, but again it appeared that linear programming rather than constraint logic programming was being used. There is some similar work (on logging for paper production) at VTT in Helsinki. There are plans to develop the system also for the Chemical Industry.
(3) Arabisation and Man-Machine Communication
(a) Automatic translation of documents (Torjoman): an excellent demonstration summarising the components described below into one system. Torjoman (developed from 1989 and ported to PC/Windows in 1994) is capable of translating not only words (with correct grammar) between Arabic and English but also handles successfully verb phrases, proper names (such as countries) and even proverbs. Behind the system are complex dictionaries (essentially thesauri since they extend beyond synonyms) but performance is impressive. This work is equal to or better than anything in EU.
(b) Natural language translation: basically components of the Torjoman technology used as an assistant, coupling the dictionaries to a helpful human-computer interface for translation assistance.
(c) Synthesis of Arab words: two projects were presented. The first demonstrated speech synthesis from Arabic script and has potential applications in Tunis Air and some banks. The second demonstrated the complex software required to manage OCR (optical character recognition) of Arabic words from Arabic script. The detailed use of height above and below the (detected) writing line, of associated diacritic marks (together giving horizontal segmentation) and of gaps in the writing line (giving vertical segmentation) with complex use of probabilistic and Markovian reasoning is very impressive. The detection of font style and size and the fallback position (should character position detection fail) of overall script word shape is practical and effective. This work is leading edge on any international scale.
Finally, it is worth noting that the IRSIT library, although small, contains the most important material (including journals) related to the fields of interest of the Institute. Clearly IRSIT is a major IT R&D site in Tunisia which has the necessary infrastructure to participate in international projects as well as competent and motivated researchers.