A logic model for model management: an embedded languages approach
Hemant K. Bhargava, Steven Orla Kimbrough · Scholarly Commons (University of Pennsylvania) · 1990
Model management research seeks to improve decision making by creating tools that facilitate the use of models, data, and other knowledge. Earlier modeling systems incorporated procedures for model solution, but did little to support many other phases in the modeling life cycle. General agreement obtains that these other phases must be supported for the use of mathematical models to become more widespread and effective. Many recent approaches to model management use an executable modeling language (EML) to represent mathematical models declaratively, and to manipulate this representation to perform functions such as model translation. This thesis extends the EML approach, arguing that existing modeling systems lack adequate means for representing, and usefully extracting, much information about models and modeling elements other than that required to just solve models. It aims to provide a formal, general means to represent and usefully extract such information, by developing the embedded languages technique and applying it to model management. In the embedded languages technique (as applied to model management) an EML, which partially formalizes the target language that we use to represent, describe, and reason about models, is embedded within a formal embedding language for model management. The technique creates dual interpretation for the symbols and expressions in the embedded language, through rules of formation such that terms and formulas in the embedded language can be treated as terms in the embedding language. This strategy enables the embedded language to (1) represent rigorously, flexibly and with generality, a rich variety of qualitative knowledge about models that could normally not be stated in the modeling language itself, (2) use this knowledge in defining inferences that are useful in the modeling process, (3) represent rules of formation for expressions in the modeling language, and use them to examine validity of model declarations, and (4) embed and integrate multiple languages. This technique was applied to develop a prototype modeling system, now in use at the U.S. Coast Guard, for the rapid development, documentation, and use of a wide class of models.