S-NET: a foundation for knowledge representation languages
Makoto Nagao, Jun’ichi Tsujii
Abstract
Makoto Nagao, Jun’ichi Tsujii
Abstract
A new knowledge representation scheme called S-Net is presented. The S-Net is a descendent of both semantic networks and recently developed AI languages. We are willing to introduce procedures into our network notations as FRL and KRL do. However, these languages have a serious disadvantage that the programmer should specify, explicitly by using indicators such as WHEN-FILLED, TO-FILL etc., when the attached procedures are invoked. This results in the restriction of system's abilities of solving problems. To avoid this, the programmer should always pay attention to the overal control issues during the coding of his procedures. In our new formalism, the explicit specification is not necessary. The problem solver based on the S-Net dynamically determines which procedure is invoked when, according to the problem solving situations, not to the pre-specified indicators. We also discuss the problems of property inheritance through hierarchy. The inheritance mechanism provided so far is so restrictive that the programmer should do all things in his procedures. In the S-Net, the property inheritance is further augmented by the 'explicit path specification'. The detailed construction of the problem solver, which performs both forward and backward reasonings appropriately, is also given.
OpenAlex reports 2 citations for this work. Citation counts describe recorded attention and do not establish research quality.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
A new knowledge representation scheme called S-Net is presented. The S-Net is a descendent of both semantic networks and recently developed AI languages. We are willing to introduce procedures into our network notations as FRL and KRL do. However, these languages have a serious disadvantage that the programmer should specify, explicitly by using indicators such as WHEN-FILLED, TO-FILL etc., when the attached procedures are invoked. This results in the restriction of system's abilities of solving problems. To avoid this, the programmer should always pay attention to the overal control issues during the coding of his procedures. In our new formalism, the explicit specification is not necessary. The problem solver based on the S-Net dynamically determines which procedure is invoked when, according to the problem solving situations, not to the pre-specified indicators. We also discuss the problems of property inheritance through hierarchy. The inheritance mechanism provided so far is so restrictive that the programmer should do all things in his procedures. In the S-Net, the property inheritance is further augmented by the 'explicit path specification'. The detailed construction of the problem solver, which performs both forward and backward reasonings appropriately, is also given.
Key concepts: Computer science, Programmer, Inheritance (genetic algorithm), Theoretical computer science, Solver, Notation, Programming language, Net (polyhedron)