Case-based Knowledge Reasoning And Its Application In Marking System

One of the main features of mental labour is reasoning.For computers,the key to have intelligence is automated reasoning.In China,a lot of remarkable achievements in automated reasoning have been made,especially in geometry of junior high school(we call it junior geometry for short). In this thesis,we are intending to achieve automated reasoning in junior geometry,based on Case-based Reasoning(CBR for short).Through conducting further analysis on the features of automated-reasoning in junior geometry and key techniques of CBR,we put forward a serials of algorithms to counter the crucial problem in CBR of junior geometry,and then we achieve a Casebased backward reasoning system in junior geometry.In the end,we apply this system to the automatic marking system.Mainly about:? Achieving knowledge representation in junior mathmatics.Acoording to the characteristic of knowledge of junior mathematics,we put forward a layered knowledge presentation of junior mathmathics.The knowledge in junior mathematics can be divided into three layers,which in order are individual,atomic proposition and instance.Individual is basic element of atomic proposition,proposition is basic element of instance.? The retrieval of instance is a very important link in CBR,it has a huge e?ect on efficiency and precision of CBR.In this thesis,we define a indexing for instance,and based on the indexing,we realized a algorithm on the similarity between a instance and current contex.Besides,we put forward strategy of structuring a case base,according to which,instances are classified based on the predicate types and stored in the instance(generally,case) base.In the end, based on the instance base and similarity,we get a e?ective instance retrieval strategy for backward reasoning.? Because the retrieved instance is inconsistent with current context,it cannot di-rectly used in current contex.So before use,we need adapt the instance to current context(we call this process homogenization).But in CBR,there is not a common approach of adaptation.We should research the adaptation respectively according the specific domains.Through in-depth study on junior geometry reasoning,we put forward a approach to get maps of points and variables from a instance to current context.Upon the maps,we implement replacement of points and variables in the instance.Thereby,we make the instance consistent with current context.? Manual entry and machine learning are two import methods of acquiring instances.But both are confronted with a problem,that is the instance acquired may be redundant for the case base.To solve this problem,we propose a redundancy detection algorithm.? Lastly,as a application of CBR of junior geometry,designing a automatic marking system for junior mathmatics.The marking on geometry steps is mainly based on backward CBR we created and the marking on algebra steps is mainly based on CAS we created.