SCHEMA: A Knowledge Edition Interface for Obtaining Program Code from Structured Descriptions of PSMs: Two Case Studies
作者:J.C. Herrero, J. Mira
摘要
The basic conjecture in this paper is that besides the basic libraries of tasks and problem-solving methods (PSMs), it is necessary to develop two complementary twin libraries. One of them consists of knowledge-acquisition schemas, as they are required by PSMs, and the other one contains those PSMs' reduction methods, from the knowledge level to the symbol level. In order to support this conjecture, we first describe the reduction method fundamentals based on hierarchical graphs representing the underlying computational model. Then we shall comment on the development of the SCHEMA interface; by using this interface, we can directly obtain the program code, provided the task knowledge, PSM, and application domain are edited following the knowledge acquisition schemas by means of structured natural language sentences. This kind of editing unmistakably and in reversibly establishes the relationships with the underlying model. Since the reduction method links the underlying model with the program code, the reduction process is completed. Conversely, we can retrieve the underlying model graph and the knowledge-level model from the program code because of the reversibility of the reduction method. In order to make clear the reduction method and SCHEMA interface functioning from the user viewpoint, we shall apply them to a classification and diagnose generic task (Hierarchical Classification), which will be decomposed by using the “Establish and Define” PSM, and another task to carry out a plan, which will be decomposed by using the “act-check-decide” PSM. We shall finish this paper with a reflection on the knowledge-level modelling and the necessity of an increase of the available reduction methods and knowledge acquisition schemas which are included in our SCHEMA interface.
论文关键词:knowledge edition interfaces, operationalization methods, hierarchical classification, protocols
论文评审过程:
论文官网地址:https://doi.org/10.1023/A:1008319718326