iAIML: Um mecanismo para o tratamento de intenÃÃo em chatterbots / iAIML: A Mechanism to Treat Intentionality in Chatterbots

AUTOR(ES)

Andrà Menezes Marques das Neves

DATA DE PUBLICAÇÃO

2005

RESUMO

Chatterbots are computational systems that aim at maintaining dialogues in natural language as if they were humans. The research work presented here had as main goal to improve chatterbotsâ performance in dialogues with users. ELIZA, the first chatterbot, was developed by Weizenbaum in 1965. Since then, several systems were developed with the aim of improving ELIZAâs performance. However, important issues are still open, among which, the treatment for ntentionality, a central point in the interpretation of human dialogue. We developed a mechanism for the treatment of intentions to be incorporate into chatterbots based on AIML. We adopted as conceptual base for this work the Conversational Analysis Theory (CAT), which regards intentions in adjacent pairs, rather than solely in the speakerâs sentence (as the Theory of the Acts of Speech). Based on CAT and on our experiments, we selected a set of intentions, that were used in the creation of AIML rules that use intentionality information to interpret and generate sentences in dialogues. The final solution was tested in a series of experiments, and it demonstrated to be capable of coping with some problems found in dialogues with chatterbots. For example, the system based on standard AIML treated 40% of the users sentences as âignoredâ, whereas our system classified only 3.5% of the sentences as totally âignoredâ. Besides, the system was able to maintain the global structure of the dialogues, criticizing opening or closing users turns typed during the development of the dialogue. Finally, we implemented three chatterbots applications, what demonstrates that the adopted solution favors the reuse of categories in AIML bases, a very expensive process from the viewpoint of software engineering

ASSUNTO(S)

chatterbot aiml ciencia da computacao artificial intelligence inteligÃncia artificial aiml chatterbo

Documentos Relacionados

• O processo da cristalizaÃÃo da intenÃÃo do voto: sob a Ãtica da Teoria dos Custos de TransaÃÃo-ECT
• Impacto dos valores organizacionais e da satisfaÃÃo no trabalho na intenÃÃo de rotatividade
• O currÃculo do ensino mÃdio integrado: da intenÃÃo à realizaÃÃo
• AvaliaÃÃo da intenÃÃo de realizar mamografia como subsÃdio para aÃÃes na promoÃÃo da saÃde
• Persona-AIML: uma arquitetura para desenvolver chatterbots com personalidade