A variety of reasoning formalisms are available which can represent the knowledge needed in text planning and sentence planning and can capture the needed inferences. The hard part is formalizing the actual knowledge needed for particular problems. A graphical language for specifying processes is a natural medium for streamlining and automating as much of this formalization as possible. There are two reasons why this is a promising area of research.
First, we can associate the constructs of a graphical language with logical descriptions automatically. This can be done across the board for simple graphical languages with precise semantics [BALC95]. It can also be done by composing programs from larger chunks whose logical descriptions are well-defined.
Second, we can easily augment the graphical language with interactive tools for specifying and editing the logical descriptions themselves. As in work like [MFCS87,EPM93], a user-interface that constrains its input to match an ontology of possible specifications can simplify the task of obtaining high-level knowledge about a process and improve the accuracy and efficiency of building knowledge-intensive systems.