A semantic-based conversational agent framework

Abstract

This paper focuses on the implementation of a novel semantic-based Conversational Agent (CA) framework. Traditional CA frameworks interpret scripts consisting of structural patterns of sentences. User input is matched against such patterns and an associated response is sent as output. This technique, which takes into account solely surface information, that is, the structural form of a sentence, requires the scripter to anticipate the inordinate ways that a user may send input. This is a tiresome and time-consuming process. As such, a semantic-based CA that interprets scripts consisting of natural language sentences will alleviate such burden. Using a pre-determined, domain-specific scenario, the CA was evaluated by participants indicating promising results.