[in Japanese]

View full abstract

When people stroll through a town, how do they select their routes and what do they look at? The route selection brings a new townscape and the new townscape attracts people toward a new direction. They induce each other, consequently. The route selection and the townscape perception of visitors who have come to the town for the first time are, however, different from those of residents who have lived there long time. Many cities put a lot of work into tourism industry by trial and error, but it is important to understand the rules of the visitors' unconscious behavior scientifically and to apply them to the tourism. This paper carries out comparisons of route selection and townscape perception between residents and visitors based on experimental data.

View full abstract

When a large-scale disaster occurs, there is a need to develop a system that can estimate amount of disaster waste, its distribution and contents, and to perform subsequent restoration activities quickly and effectively. We are sure that estimating disaster waste using remote sensing data is the first priority to affect all subsequent processing. However, it is necessary to develop a method for extracting detailed information buried in mixels of remote sensingdata. In this paper, we proposed a method for unmixing mixels that uses the Digital Number (DN) and texture features appeared in the THEOS data.

View full abstract

Train plays an important role as public transportation in our daily life. However, if transport volume decreased due to irregularities in the timetable of train, it'll be shock to our society. In that case, it is needed for train operators to recover the transport volume by their operating skills. In this study, we considered operation of expert operators and developed an intelligent operating support system which is considering the change of constraint conditions based on the knowledge of expert operators in the dynamic environment. By providing appropriate operating support information to operators, it is confirmed that the developed intelligent operating support system is effective for shortening arrival time and the recovery of transport volume in such environment.

View full abstract

Mathematical programming has been applied to many problems in various fields. The data of real problems contain uncertainty and are thus represented as random variables. Decision-making under conditions of uncertainty involves potential risk. Stochastic programming deals with optimization under uncertainty. A stochastic programming problem with recourse is referred to as a twostage stochastic problem. In this paper, the stochastic programming model for the logistics network reorganization problem and the efficient solution method are shown. The traditional expected cost model and the CVaR model are compared in numerical experiments. The expected cost rises slightly using the CVaR as the objective. However, it is shown that the worst cost is reduced using the CVaR.

View full abstract

Psychological and physiological effects of colored lights for change of flashing cycle were examined using aliquid crystal display in a dark environment (<0.1lx). The colored lights were characterized with various flashing cycle, i.e., flashing cycle were selected the range of 1.2-4.2 s. To investigate psychological and physiological effects, we used electroencephalography, heart rate and a simple questionnaire. Subjective evaluations of 'comfort', 'relaxation', 'awakening', 'tension', and 'anxiety' were conducted using a five-level psychological preferences evaluation. Electroencephalographic, heart rate and psychological evaluations revealed that the slower flashing cycle had a good effect for relaxation. Colored lights with the earlier flashing cycle enhanced the degrees of tension and anxiety. Results of this study are effective for providing healing environment to patients with severe stress.

View full abstract

We propose a support system using question answering technology for reading of novels on text files. The system outputs the candidates of the answers to questions from the user about the content of the novel. The candidates are searched from the text of the novel and sorted based on the evaluation value. The system is equipped with several additional functions such as "Spoiler Preventer", "Character Information Summarizer" and "Progression-aware Outline Function". In this paper, we report the system overview, implementation and examples of execution. We carried out an experiment for the evaluation of the system and proposed method. Ten subjects used the system and we found that they accepted the proposed method from the results of the experiment.

View full abstract

Statistical studies in linguistics have shown that relations between word frequencies and their ranks have the regularity of power-law distributions. However, the self-organizing behaviors of word occurrence in the relations have not yet been studied comprehensively. We carried out an empirical analysis of medical term occurrences using the Merck Manual, which is a well-known medical reference book. We investigated medical term connectivity by modeling term occurrences in a text as state transitions in a system. Observing connection frequency distributions, we then found scale-free patterns of connectivity such that frequencies of both incoming and outgoing connections of each term would be power-law distributed. Our results suggest that the term connectivity may give the auto-regulatory dynamic behavior of self-organization. Investigating term connectivity dynamics may significantly advance text processing methodologies.

View full abstract

Large amounts of good quality parallel texts (translation memory) are required for statistical machine translation methods. While collecting large amounts of parallel texts has been studied so far, a methodology of managing qualityby revising sentences for parallel text improvement (clean-up)has not been well-known. Here parallel text ranking based on quantification about word alignments is described. We show ranking experimental results for the clean-up using about 400 thousand parallel Japanese-English sentences.

View full abstract