الفهرس 
Natural intelligenceOnly 14 pages are availabe for public view
 |  | from | 204 |  |  |
| | | from | 204 | | |
Abstract
The thesis intended to develop and implement a new approach for measuring the Intelligence Ratio (IR) inside an Expert System (ES). The proposed approach chooses only two main parameters that affect the General Intelligence Ratio (GIR) inside an ES. These two parameters are the knowledge acquisition (KA) and reaction time (RT). Many related works have been done since the 1950s when, the greatest English mathematician, Turing proposed his famous Turing test to determine whether or not a computer program is intelligent. Another trial to determine the machine intelligence ratio have been done by Slagee (1988), Preece (1990) through evaluation of ES performance. Such trials of measuring the IR inside any AI program will address many topics that improve their performance and hence increase the level of developing an AI program especially ESs towards the knowledge revolution. Researchers could also use it to better understand how human reason. Success in this endeavor would provide a great benefit to mankind. In order to accomplish the measurements of IR, four expert systems have been designed. The four expert systems are in the chemical analysis domain. These systems are designed in such way that each one has the ability to detect Unknown Chemical Radicals (UCR) in different manners to capture the effect of the main two parameters of IR; knowledge acquisition and the reaction time. The result of the performance of the four systems are calculated and compared. The comparison is applied on 20 UCRs (acidic and basic). The Processing Time (PT) for each system is calculated. For more accuracy, the average of PTs (APT) over the 20 UCRs also is calculated for each system. Also the average of the number of the asked questions (ANQ), for detecting any UCR, is calculated. Moreover the parameter of knowledge acquisition between the four systems is compared. The systems tests are executed under a normalized environments; the same knowledge base, the same computer [P III, 800 MHZ, 64 MB RAM, and WINDOWS 2000], and the same formal language [PROLOG] platform. According to the obtained results, the first expert system has not any level of intelligence because it has not any knowledge acquisition ability. The second, the third, and the fourth systems give certain level of intelligence because of their knowledge acquisition ability. The fourth system design has an organized Knowledge Base (KB) than the second system and the third system. Therefore the fourth system results are more logic and it is more intelligent than the other two systems (the second one and the third one) for detecting any unknown radical. It is found that the proposed approach for measuring the IR of the designed systems is acceptable and could be applied for other expert systems domains.