The conversational knowledge process is a collective activity for knowledge creation, management, and application where conversational communications are used as a primary means of interaction among participating actors. In this article, I survey a suite of communication tools that augment the conversational knowledge process in order to accelerate social problem solving.

抄録全体を表示

Today it is pointed out that social acceptance of new science & technology should be decided not only by the experts but also by citizens, because new science & technology gives great affects on the daily life of citizens. This paper deals with information systems to support communication for citizens. Here the communication for citizens is divided into two aspects of communication, dialogue with citizens and information presentation to citizens. First, the problems of communication for citizens are analyzed and the methods and approaches to cope with the problems are investigated. Then, research projects to enhance this type of communication are introduced concretely. We believe the importance of communication support for citizens will be increased in near future.

抄録全体を表示

This study presents a simple method for examining a child's development by automatically analyzing the child's reactions while feedback images are displayed for 2 min. This computer analysis recognizes some features of an image of the child's reactions. This system can perform that analysis in a shorter time than conventional methods, as typified by the Enjoji-method. Moreover, evaluation results of our system verify its validity compared to conventional methods. This system is applicable to achieve automatic analyses through a network to thereby realize telemedicine as part of a Multimedia Communication System.

抄録全体を表示

This paper presents experiments conducted to evaluate an automatic video editing system, founded on vision-based head tracking, that clearly conveys face-to-face multiparty conversations, such as meetings, to viewers. Systems that archive meetings and teleconferences to effectively facilitate human communication are attracting considerable interest. Conventional systems use a fixed-viewpoint camera and simple camera selection based on participants' utterances. Unfortunately, they fail to adequately convey who is talking to whom and nonverbal information about participants etc., to viewers. To solve this problem, we previously proposed an automatic video editing system using vision-based head tracking. This paper describes subjective evaluation experiments in which videos of entire conversations with 3 participants were presented to viewers; the results confirm the effectiveness of our system.

抄録全体を表示

In order to make collaborative business activities fruitful, it is essential to know characteristics of organizations and persons in more details and to gather information relevant to the activities. In this paper, we describe a notion of “information recycle" that actualizes these requirements by analyzing documents. The key of recycling information is to utilize annotations on documents as clues for generating users' profiles and for weighting contents in the context of the activities. We also propose a method of extracting annotations on paper documents just by pressing one button with the help of techniques of camera-based document image analysis. Experimental results demonstrate that it is fundamentally capable of acquiring annotations on paper documents on condition that their electronic versions without annotations are available for the processing.

抄録全体を表示

A multiagent system (MAS) has recently gained public attention as a technique to solve competition and cooperation in distributed systems. However, MAS's vulnerability due to the propagation of failures prevents from applying to a large scale system. Then, this paper proposes a general composition technique to improve its reliability easily applied to the existent MAS. Proposed system monitors messages between agents to detect undesirable behaviors (failures). Collecting related information, the system generates global information of interdependence between agents and expresses it in a graph. This interdependence graph enables us to detect or predict undesirable behaviors. This paper also shows that the system can optimize performance of MAS and improve adaptively its reliability under complicated and dynamic environment by applying the global information acquired from analysis of the interdependence graph to a replication system.

抄録全体を表示

The financial service counter is consisted of several advisory services, and service improvement is needed to offer effectively proper responses for various kinds of customer request. But it is not practical to put various expert staffs at each financial branch office, because of much cost and task performance falling. This paper proposes the window consultation system with contact persons at the financial branch office through on-demand calling of a proper expert for a problem. This proposed system has mainly two functions, remote experts selection and tele-consultation function, to offer replies or expert advises to the problem, using multimedia network technology. Customers can consult effectively with a proper expert staff about their requests through the proposed system, because it can connect the expert by the clerk of the window counter who comprehends beforehand the customer's requests. This system with contact persons can offer high quality services and avoid mistakes, which a customer makes in operating the consultation systems terminal alone by oneself. And this system is composed of three-layered structures, application component layer, middleware component layer and fundamental component layer. This structure is able to construct much easily various services at various scale financial branch office.

抄録全体を表示

Using a head-up display (HUD) for computer-assisted instruction (CAI) creates a new form of support for manipulating complicated devices. A head-mounted display (HMD) enables head-up interaction to be realized by overlaying virtual objects onto the real scene. We attempted to apply HUD-based CAI to a support system for the operation of a transportable earth station consisting of an automobile van containing various pieces of equipment for satellite communications. The HUD-based CAI system provides information on how to manage the pieces of equipment, which are identified with markers, as the user looks at them, which enables hands-free presentation. Although many HUD-based systems have been developed for proof-of-concept or demonstration purposes, there have been few user studies on the practical use of HUD-based systems in CAI. We examined a practical CAI system used to manipulate the equipment in the transportable earth station. We first report on a preliminary experiment to investigate the readability and operability of HMDs. A single-sided HMD outperformed a two-sided HMD with regard to transcript tasks. In addition, we performed a usability study to compare the HUD-based CAI system to conventional printed materials and laptop PCs. The result of the operation tasks showed the better performance for the CAI system than that for the other media, and suggested that the CAI system improved reliability of information acquisition for both the instructions and the equipment.

抄録全体を表示

In this paper, we propose a unified synchronous/asynchronous distance learning support system using real-time editing method of teaching materials. In the real-time editing method, teachers first create teaching-material files to which the content elements constituted for every lesson subject are classified and registered. Next, teachers input editing commands on the internet during the lecture and edit teaching materials on real time with editing software, and transmit them to remote students. By sending the teaching material files to the student side beforehand, and performing teaching-materials editing software on the student side, it becomes possible to display teaching materials on the student side only by the command input from the teacher even if in the low speed network. In this way, teachers are able to reflect students' questions of an asynchronous manner and reactions during classes into editing of teaching materials. In this paper, we describe details of the proposed method and the results of the experimental evaluations.

抄録全体を表示

This paper describes how to reuse a series of ad-hoc tasks such as special meeting arrangement and equipment procurement. Our RESTER (Reusable TODO Synthesizer) allows a group to reuse a series of tasks which are recorded in case database. Given a specific event, RESTER repairs the retrieved similar case by the ontology which describes the relationship of concept in the organization. A user has chance to check the modified case and to update it if he finds that there are incorrect repair because of deficient ontology. The user is also requested to judge if the retrieved case works or not. If he judges it is useful, the case becomes to be reused more frequently. Thus, RESTER works under the premise of human-computer collaboration. Based on the presented framework, this paper has identified several desirable attributes: (1) RESTER allows a group to externalize its experience on jobs, (2) Externalized experience are connected in case database, (3) A case is internalized by other group when it is retrieved and repaired for a new event, (4) New job generated from the previous similar job of one group is socialized by the other group.

抄録全体を表示

We propose an improvement of an information filtering system with independent components selection. The independent components are obtained by Independent Component Analysis and considered as topics. It is effective for improving accuracy of information filtering to select some similar topics by focusing on these meaning. To achieve this, we select the topics by Maximum Distance Algorithm with Jensen-Shannon divergence. In addition, document vectors are represented by the selected topics. We create a user profile from transformed data with a relevance feedback. Finally, we recommend documents by the user profile and evaluate the accuracy by imputation precision. We carry out an evaluation experiment to confirm availability of the proposed method and also consider the meaning of components in this experiment.

抄録全体を表示

Data retention, regulatory compliance, and security requirements are increasing by explosive information growth and government regulations. Information Lifecycle Management (ILM) has gotten a lot of attention as a solution for these requirements recently. The goal of ILM is to ensure that information is stored on the most effective and appropriate storage medium depending on the service level required in the phase of the life of information. And the goal of our research is to automate the ILM process using “policy" indicating managers' intension. This paper proposes the method for description and interpretation of the policy.

抄録全体を表示

To dematerialize the securities such as stocks or cooporate bonds, the securities were registered to account in the registration agencies which were connected as tree. This tree structure had the advantage in the management of the securities those were issued large amount and number of brands of securities were limited.
But when the securities such as account receivables or advance notes are dematerialized, number of brands of the securities increases extremely. In this case, the management of securities with tree structure becomes very difficult because of the concentration of information to root of the tree. To resolve this problem, using the graph structure is assumed instead of the tree structure. When the securities are kept with tree structure, the delivery path of securities is unique, but when securities are kept with graph structure, path of delivery is not unique. In this report, we describe the requirement of the delivery path of securities, and we describe selecting method of the path.

抄録全体を表示

Texture segmentation methods using split and merge, pyramid node linking, or active contour model, et al. has been proposed. However, they need to tune parameters for the optimum segmentation of every image querying. It is difficult to fix a parameter set for all the query images. To solve this problem, a powerful method using simple genetic algorithm (SGA) is proposed. Segmentation results do not vary according to a tuned parameter set for the evolutionary algorithm. But, since it takes the global and parallel optimization approach on an image plane, it cannot search the optimum solution efficiently.
In this paper, for efficient optimization, we propose a new evolutionary segmentation method of texture image using perturbation optimization. After dividing the original image into small rectangular regions, we extract feature vectors from small regions. We search the candidate of cluster numbers for each small region by perturbating the cluster number into another one and by inspecting the change of the fitness for each perturbed number. The combination of cluster numbers is optimized locally and successively on an image plane by the perturbation optimization. This brings us the enhancement of segmentation accuracy and the curtailment of processing time compared with the method using SGA.

抄録全体を表示

The template matching is a method to search a target image area that is similar to a template image. However, the target image area must have the right position as the template image generally. Therefore it requires large computational cost when a location and an inclination of the target image area are unknown. This paper proposes a method to search a target image area that has a free location and a free inclination by geometrical parameters obtained by equidistantly edge group. In this method, the difference of directions is calculated by the edge direction of a referenced edge pixel and by the directions from the referenced edge pixel to other selected edge pixels. After that, the matching process is performed by using the difference of directions and matching rates are accumulated. Then the location and inclination of the target image area are searched by scanning the accumulation space. The experimental results show that the proposed method had higher reliability and less computational cost for image matching.

抄録全体を表示

A quantum mechanical wave function is shown to approximate affine nonlinear optimal feedback, when an absolute value of the terminal wave function is set positive and with no singular dependence on a control constant H_R. The wave function is explicitly written down according to a path integral method and a stationary phase approximation of the integral. It is shown that a phase of the wave function approximates in H_R → 0 to a Hamilton-Jacobi value function. We can phase the quantum mechanical fluctuation out in the limit. It is simple to take the terminal absolute value function that meets the condition of having no singularity at H_R=0. This is because the terminal absolute value function without any dependence on the constant H_R apparently satisfies the no singularity condition. Although we restrict ourselves with a scalar system, generalization to a system with higher dimensionality is straightforward.

抄録全体を表示

This paper proposes a novel algorithm for iterative tuning of a completely unknown two-mass torsional motor system. The command generator tracking (CGT) theory is used to formulate a feedforward controller that consists of a reference model and input generator, which can clarify the relationship between the controller parameters and the two-mass model parameters. The relationship is used to construct physical model parameter estimates, which are shown to converge to true values as the output error is minimized. The proposed tuning process is conducted by utilizing these estimates at each iteration to approximate a system transfer function. This allows the analytical calculation of controller gradients needed for conducting Iterative Feedback Tuning (IFT), without having to perform additional experiments for gradient estimation. The effectiveness of the proposed method in achieving trajectory tracking and identification of all physical parameters is demonstrated via experimentation.

抄録全体を表示

Independent component analysis (ICA) is a technique of transforming observation signals into their unknown independent components; hence ICA has been often applied to blind signal separation problems. In this application, it is expected that the obtained independent components extract essential information of independent signal sources from input data in an unsupervised fashion. Based on such characteristics, ICA is recently utilized as a feature extraction method for images and sounds for recognition purposes. However, since ICA is an unsupervised learning, the obtained independent components are not always useful in recognition. To overcome this problem, we propose a supervised approach to ICA using category information. The proposed method is implemented in a conventional three-layered neural network, but its objective function to be minimized is defined for not only the output layer but also the hidden layer. The objective function consists of the following two terms: one evaluates the kurtosis of hidden unit outputs and the other evaluates the error between output signals and their teacher signals. The experiments are performed for some standard datasets to evaluate the proposed algorithm. It is verified that higher recognition accuracy is attained by the proposed method as compared with a conventional ICA algorithm.

抄録全体を表示

Recently, Genetic Network Programming (GNP) has been proposed as one of the evolutionary algorithms. It represents its solutions as directed graph structures and the distinguished abilities have been shown. However, when we apply GNP to complex problems like the real world one, GNP must have robustness against the changes of environments and evolve quickly. Therefore, we introduced Automatically Generated Macro Nodes (AGMs) to GNP (GNP with AGMs). Actually GNP with AGMs has shown higher performances than the conventional GNP in terms of the fitness and the speed of evolution. In this paper, a new mechanism, AGMs with variable size, is introduced to GNP. Conventional AGMs have the fixed number of nodes and they evolve using only genetic operations, while a new method allows AGM to add nodes by necessity and delete nodes which do not contribute to the evolution of the AGM. The proposed GNP with AGMs of variable size is expected to evolve effectively and efficiently when it is applied to agent systems and also expected to make better behavior sequences of agents more easily than the conventional GNP algorithm. In the simulations, the proposed and conventional methods are applied to a tileworld problem and they are compared with each other. From the results, GNP with AGMs of variable size shows better fitness than GNP with AGMs of fixed size and the conventional GNP when adapting ten different environments.

抄録全体を表示

This paper proposes a method of predicting which customers' account balances will increase by using data mining to effectively and efficiently promote sales. Prediction by mining all the data in a business is difficult because of much time required to collect, process, and calculate it. The selection of which features are used for prediction is a critical issue. We propose a method of selecting features to improve the accuracy of prediction within practical time limits. It consists of three parts: (1) converting collected features into financial behavior features that reflect customer actions, (2) extracting features affecting increases in account balances from these collected and financial behavior features, and (3) predicting customers whose account balances will increase based on the extracted features. We found the accuracy of prediction in an experiment with our method to be higher than with other conventional methods.

抄録全体を表示

Wrong:were also evalutated
Right:were also evaluated

抄録全体を表示

Wrong:with those charecteristics
Right:with those characteristics

抄録全体を表示

Wrong:psychosmatic state
Right:psychosomatic state

抄録全体を表示

Wrong:since the constructed T-S Model
Right:since the constructed T-S model

抄録全体を表示

Wrong:how to construct an T-S model
Right:how to construct a T-S model

抄録全体を表示

Wrong:an optimal discriminat function
Right:an optimal discriminant function

抄録全体を表示

文献(1)

抄録全体を表示

文献(4)

抄録全体を表示

Wrong:we comfirm
Right:we confirm

抄録全体を表示

文献(1)

抄録全体を表示

文献(6)

抄録全体を表示