Emotion and automatic dialogue summarisation / Emoção e a sumarização automatica de dialogos

AUTOR(ES)

Norton Trevisan Roman

DATA DE PUBLICAÇÃO

2007

RESUMO

This thesis presents a number of contributions to the field of automatic dialogue summarisation. It provides evidence for the hypothesis that whenever a dialogue features very impolite behaviour by one or more of its interlocutors, this behaviour will tend to be described in the dialogue s summary. Moreover, further experimental results showed that this behaviour is reported with a strong bias determined by the point of view of the summariser. This result was not affected by constraints on the summary length. The experiments provided useful information on when and how assessments of emotion and behaviour should be added to a dialogue summary. To conduct the experiments, a categorical multi-dimensional annotation scheme was developed which may also be helpful to other researchers who need to annotate data in a similar way. The results from the empirical studies were used to build an automatic dialogue summarisation system, in order to test their computational applicability. The system s output consists of summaries in which technical and emotional information, such as assessments of the dialogue participants behaviour, are combined in a way that reflects the bias of the summariser, being the point of view defined by the user.

ASSUNTO(S)

artificial intelligence behavioral analysis inteligencia artificial natural language processing (computer science) processamento da linguagem natural (computação) comportamento - avaliação

Documentos Relacionados

• Automatic detection of significative events and summarization in soccer transmissions
• Dialogue and Reflection
• SABIO : abordagem conexionista supervisionada para sumarização automatica de textos
• CATARSE, EMOÇÃO E PRAZER NA POÉTICA DE ARISTÓTELES
• Sumarização automática de rushes vídeos baseada em características espaciais e espaço-temporais