Vega - An Intelligent System Development Tool

Štefan Furlan (2006) Vega - An Intelligent System Development Tool. Prešeren awards for students.

Full text not available from this repository.

Abstract

There are a lot of problems which computers solve quicker and better than humans, but there is also a varietz of problems at which humans are substentially better. For solving such problems we use intelligent systems. The intelligent system lifecycle consists of three basic phases, which are knowledge acquisition phase, the same does not hold for the other two phases. Even more, the developers are doing a lot of unnecessary work in those phases, repeating the work, which has already been done successfully. That causes a longer development time and poor-quality products. The aforementioned deficiencies can be eliminated to a certain degree, using approaches that are already well established in the field of object tehnologies. Knowing that, led us to design and construct a tool, which uses these approaches to support the construction and use phases of inteliigent systems' life cycle. The tool was named Vega, after the renowned Slovene mathematician. In the construction phase, Vega allows knowledge worker to integrate the knowledge, obtained in the knowledge acquisition phase, into a ready-to-use intelligent system. In the later phase Vega acts as a host, allowing outer entities to access adn execute hosted intelligent systems. The user can access the intelligent systems via different user applications, whereas other applications can make use of the same functionalities via application interface. It has a form of a web service and as such enables platform independency. There is no well-known system of criteria, upon which Vega could be evaluated. Therefore we decided to try it in practice. The case included a design and implementation of threeintelligent systems in Vega to assist BITOLA system at several tasks. BITOLA is a discovery support system. Its discovery algorithm searches for not yet discovered links between different biomedical terms via text mining scientific articles database. For more accurate results the concepts are limited to the gene region. The gene symbols, used in this process, are very ambigious and therefore some of the results are always useless. We developed three intelligent systems in Vega that assist BITOLA at the task gene symbol disambiguation and therefore improve the quality of the BITOLA's services. Using Vega, the whole intelligent system development process was simple and therefore faster. On that basis we are able to conclude that the use of Vega had paid out.

Actions (login required)