A Compiler-Interpreter-System for Decoding the User's Intention Within a Speech Understanding Application

For a speech understanding graphic editor, a compiler-interpreter-system is introduced to process a semantic structure, a special form for representing the semantic content of a spoken utterance. After a semantic structure has been converted to database queries by the compiler, the interpreter processes all these queries and updates the corresponding database, i.e. in our case the graphics data base containing the features of all objects on the screen. The system has been tested with 1843 semantic structures of spoken utterances within the ’graphic editor’ domain. The rate for correct database queries is 97.3%. Storage of the whole domain-specific knowledge in external and editable files enables easy portability to other domains.