Abstract:
In today’s IT dependent society, knowledge/experience sharing and reusing previously stored experiences in different forms are becoming more and more important due to lots of information and technical knowledge that we store daily. Case Base Reasoning (CBR) imitates such tendency of human experience based problem solving habit and it’s a methodology as how human reuse their experience.
This research study presents the possibility of using a potential CBR systems provide in order to solve complex discipline decision making in Ethiopian Higher education sectors. A decision support prototype system is developed using python general purpose programming language which imitates the process of jCOLIBRI CBR system development; the development process include PreCycle module for Text preprocessing and indexing and representation of the unstructured cases which is done using flat representation technique. The Cycle module has the pillars of the CBR cycle model the retriever, the reuse, revise and retain processes. The Searcher is done following the vector space modeling for case retrieval using cosine similarity measurements and the other processes such reuse, revise and retain are merged together and implemented as Adaptation process supported by simple reuse and modification algorithm.
And the developed Application of Amharic Text Based Reasoning for Academic Discipline Cases Decision Making and Support System is evaluated by using two separate evaluation methodologies; experts’ validation of the prototype by visual interaction with the system by selected user queries and technical verification using standard IR system evaluation methodologies. The evaluation is done using both approaches have shown better performance and suggested the commence ability of such systems in the studied domain area.
The overall system achieved great performance but there are still areas of the system that should be improved. One is that comparing two cases in free text format is a challenging task, so integration of Word Sense Disambiguation and efficient Word-Net need to be constructed. Secondly, Textual CBR adaptation of texts to facilitate machine learning process through Pseudo feedback shall be seen as alternate method and the use of auto case builder may solve the problem of the actual machine learning. Finally other SM algorithms also need to be tested for comparison and selection of better algorithms.