Key points are not available for this paper at this time.
The experimental EPISTLE system is ultimately intended to provide office workers with intelligent applications for the processing of natural language text, particularly business correspondence. A variety of possible critiques of textual material are identified in this paper, but the discussion focuses on the system's capability to detect several classes of grammatical errors, such as disagreement in number between the subject and the verb. The system's error-detection performance relies critically on its parsing component which determines the syntactic structure of each sentence and the grammatical functions fulfilled by various phrases. Details of the system's operations are provided, and some of the future critiquing objectives are outlined.
Miller et al. (Thu,) studied this question.