CONTROL SEARCH USING RULES, TESTS AND MEASURES

Publication Type:

Conference Paper

Source:

SMC Conference 2004 (2004)

URL:

files/proceedings/2004/P9.pdf

Abstract:

Problem solvings may be viewed as search. In a search, know- ledge plays a crucial role in guiding the search to an acceptable solution. We believe that an explicit separation between know- ledge levels and the use of knowledge in problem solving al- lows knowledge-rich systems to be more flexible and powerful. With this aim, appropriate representation framework and infer- ence mechanism are required. This paper discusses the issues involved by viewing ef- fective problem solving through the search control perspective. In this view, the data stream is defined as a stream of problem states returned after applying a control definition to a problem state ( ). The control definition is a control block constructed from three primitives, namely rules, tests and measures. A problem is said to be solved if a sequence of control definitions (control stream) applied to the start state yields the goal state e.g. . The stream of control procedure utilises knowledge to decide how to move from the start state to the goal state. We explain the search mechanism and how the knowledge can be applied to guide search in this view. We illustrate the approach with a case study in the harmonisation of Bach’s chorales.