This paper proposes an integrated framework of narrative discourse system that trans- forms a story text to a variety of discourse texts. This mechanism is one of the major parts of the architecture of our narrative generation system. A “story” means the content of a narrative and is described with the conceptual representation of temporal ordered events in the narrative generation system. On the other hand, a “discourse” means how to form a story and is described with the conceptual representation that is corresponded to a real text structure. In the research of narrative generation system, few approaches deal with the aspect of narrative discourse. A distinguished characteris- tic in this system is to use two literary theories for developing its important mechanisms. First, narrative discourse techniques for manipulating narrative discourse structures are defined according to the narrative discourse theory by Genette. Second, the circula- tive generation process for these narrative discourse techniques is controlled using the repetitive interaction between a narrator and a narratee based on a computational in- terpretation of the reception theory by Jauss. In the first part, this paper explains the overview of the above literary theories. And in the next parts, we describe the pro- posed system's mechanism and implementation, and several experiments of discourse generation by the system.
This paper aims to study the creative process of collaborative improvisation, which has rarely been the subject of cognitive science research. The special focus is on how a dancer’s improvisational process is influenced by another dancer’s performance. Bat- tle performances of street dance are chosen as the target performances of this study and contrasted with solo performances, because in battle situations dancers have to generate their original expressions in a highly improvised manner. We conducted an experimental study, aiming to investigate both the cognitive process of dancers and the characteristic features of their dance performances. The three main results are as follows: 1) The generation frequency of new dance movements was not different be- tween solo and battle dance performances. 2) The generation process of a dancer’s new dance movements was influenced by another dancer’s performance. In the battle per- formances, dancers generated new dance movements by combining several movements that they had already acquired, while they generated new movements by only changing some parts of their acquired dance movements in the solo performances. 3) In the battle performances, as distinct from the solo dance performances, dancers generated their dance movements by adopting movements of another dancers and changing the structures of their movements. These features resulted in differences in the processes of generating new patterns.
This study, at first, constructs a computational model of inductive reasoning based on the probabilistic concept structure estimated by the statistical analysis of large scale Chinese language data. In order to examine the efficiency of the model, which has al- ready been certified about the Japanese language (Sakamoto & Nakagawa 2008, 2010), the study verifies the validity of the model using the psychological experiment. The new computational model of inductive reasoning is constructed based on the statistical analysis of extended Japanese language data, including not only the news paper articles but also literature. The validity of the model is then verified using the psychological experiment. Furthermore, from the comparison between simulation results of both models, the study examines the hypothesis that the inductive reasoning process does not necessar- ily depend on the individual language system. Finally, through the detailed comparison between the results of both models, the commonality and difference between both cul- tures and social systems hidden in the back of both languages is discussed.
While reading silently, we process only the visual information of the text. Conversely, while reading texts orally, we process the visually presented text and produce the au- ditory information of the text through articulatory movement. These activities are assumed to facilitate the memory and comprehension of textual information. Although we cannot use such auditory nor motor information while reading silently, there is little difference between the degree of comprehension based on silent and oral reading for adult readers. The purpose of this study is to explain how we compensate the loss of multisensory process during silent reading by comparing the visual processing process during silent and oral reading. In Experiment 1, we measured the eye-movement dur- ing reading garden-path and normal sentences silently and orally. In Experiment 2, we compared the eye-movement during reading more common paragraph silently and orally. The results showed that silent reading took shorter time for comprehension than oral reading, and readers had more visual fixation points and read back frequently dur- ing reading silently than orally. These reading strategies during silent reading seemed to compensate the loss of multisensory process and support the text comprehension.