Hybrid Similarity Measure in SmartLP: A Case Based System in Lesson Planning
Keywords:case retrieval, computational approach, representational approach, similarity measure, weighting, terms expansion, case search
Planning for teaching imposes a significant burden on teachers as they need to prepare different lesson plans for different classes according to various constraints such as students’ ability and previous knowledge. SmartLP, a case-based lesson planning system, has been implemented as a means of assisting teachers in constructing quality lesson plans more quickly and efficiently. SmartLP applied Case Based Reasoning (CBR) approach which consists of the following processes; retrieve, reuse, revise, and retain within its cycle. CBR is the process of solving new problems based on the solutions of similar past problems. The first and most crucial step of solving new problem in CBR is the process of retrieval relevant cases from the case base. There are several factors that affect case retrieval such as similarity measure, indexing approach and all techniques within the approaches. Both computational (measures of similarity) and representational approaches were used for case retrieval in SmartLP which is done via lesson plan searching. There are five types of search in SmartLP; Advanced Search, Hierarchical Search, Search by Browsing, Boolean and Basic search. In Advanced Search, a hybrid approach; combination of distance based (computational) and representational was applied for case retrieval. In addition, each indexed element is assigned with certain number to indicate their relative importance as a terms weighting mechanism. Terms expansion within semantic approach was implemented in Hierarchical Search, based on lesson plan ontology. In Search by Browsing, the search terms are organised in general to specific manner which also derived from lesson plan ontology. Boolean search applied Boolean concept while Basic Search is a general search with exact search keywords. This paper presents the retrieval methods in SmartLP via five types of search.