CONVENTION AND INTENTIONALITY IN PROCESSING SIMPLE CONVERSATIONAL SCRIPTS

INTELLIGENT INTERFACE AGENTS GREATLY INCREASE THE SUITABILITY OF COMPUTER TECHNOLOGY FOR USE BY THE BROADER POPULATION IN PERFORMING SIMPLE, HUMAN-ORIENTED TASKS. EFFECTIVE INTELLIGENT AGENTS CAN BE BUILT USING MEMORY-BASED PROCESSING. THIS RESEARCH INVOLVES THE APPLICATION OF MEMORY-BASED CONVENTION, INTENTION, AND DOMAIN KNOWLEDGE TO HANDLE SIMPLE SCRIPTS ACCESSIBLE FROM THE TELEPHONE. THE RESEARCH CONCENTRATES ON RECORDING AND INTEGRATING THIS KNOWLEDGE TO PROVIDE AN INTELLIGENT AGENT FOR THE VOICE DIALING TELEPHONE SERVICE. CONVENTION AND DOMAIN KNOWLEDGE ARE REPRESENTED IN CONVERSATIONAL SCRIPTS. INTERACTIONS WITH THE SYSTEM ARE STORED IN DYNAMIC MEMORY BOTH AS LEXICAL SENSORY DATA AND EPISODIC CONCEPTUAL STRUCTURES. THESE INTERACTIONS INCLUDE TYPED NATURAL LANGUAGE, AND TYPED AND KEYED DTMF AND SILENCE TOKENS. WHEREVER POSSIBLE, EXISTING MODULES AND REPRESENTATIONS ARE BEING USED FOR PROCESSING NATURAL LANGUAGE, DYNAMIC MEMORY, REASONING, AND CONVERSATION CONTROL.