Study Of Artificial Intelligent Based Human-Computer Interface To Model Base

Human-computer interface to model base is a very significant but also toughproblem in-the fields of management decision and the development of intelligentsystem. As one part of "Management Decision-Oriented Intelligent Systems Study ofIsomeric Knowledge Representation and Knowledge Management" sponsored byNational Natural Science Foundation of China (No.70171002) , this paper is aimingat promoting the development in the study of intelligent human-computer interface tomodel base.As the core of Intelligent Decision Support System, a model system draws muchattention of scholars and experts both in academic and engineering field. A modelsystem, which builds the bridge between the decision problems and related models tokeep the details of technological realization from the decision maker, takesresponsibility for the storage,running and maintenance of the models, and offering avariety of models in a receptible and flexible way as well. However, compared withthe hot topics of model manipulation in model base management, few researcheswork on the theoretics or application of human-computer interface to model base.Moreover, it could be hardly to find any researches on intelligentization ofhuman-computer interface.The only several interface systems to model base nevercover the entire application and just the partial functional realization such as modelclass selection or model structure selection instead.Contrasted sharply with the human-computer interface to model base, theresearch on intelligentization of human-computer interface to database combinedwith artificial mtelligence approaches has greatly advanced these years, andfurthermore the natural language interface to database (NLIDB) achieves lots ofsuccesses on its application area with the realizable and confirmable performance. Therefore, the paper proposes a new intelligent human-computer interface to modelbase which integrating artificial intelligence into decision support system, as callednatural language interface to model base (NLIMB), aimed at promoting the researchon intelligentization of human-computer interface to model base and offering thedecision makers a flexible and convenient way of human-computer interaction.NLIMB could release the decision makers from being familiar with the modelclass, structure, applicable range and computational realization of varieties of modelsin model base, and transfer their concentration to the decision problem itself. NLIMBfirst processes natural language parsing of the problems inquired by decision makerby applying N-gram and Probabilistic Context-Free Grammar model intopart-of-speech tagging and syntactic parse, and then processes the potential semanticparsing. In the second stage of automatic model selection, it firstly selects the modelclass by using the hybrid inference framework proposed in this paper whichcombines rule-based and case-based reasoning; secondly selects the model structureby using Elman neural network revised in this paper which employs Gauss lostfunction as error function; and last confirms the model case by parameters estimation.NLIMB finally returns the computational result of the selected model to the decisionmaker.The limited application of human-computer interface to model base inforecasting models area, the paper applies it into business inventory decisionmanagement to demonstrate the efficiency of NLIMB, and also extend theapplication area of NLIMB. The experiment shows that NLIMB works quite wellwith the decision problems and holds the advantages of high efficiency, intelligentinteraction and flexible usage over other human-computer interface to model base.This paper studies a lot both on theory and experiments. And the researchaccords with trends of intelligent system development that has great significance onacademic theory and actual practice.