Intelligent Interactive Multimedia: Systems and Services

Recently, an intelligent robot is expected to operate our living area. To realize, a robot should make a decision for action from a simple order by human. For decision making of a robot, it is important that a perception of environmental situation and adaptation to the preference of a person. We have proposed the learning method based on SOM to adapt environmental situation and the preference of human. Through experiments by simulation, we verified that the proposed method can consider the changing of attribution by time variation. And, the decision making of a robot can be adapted to the preference of a person through interaction with a person.

Abstract. Focus+Glue+Context map system EMMA (Elastic Mobile Map) is composed of the areas of expanded detail (Focus), peripheral areas (Context), and areas that absorb the distortion between the Focus and the Context areas (Glue). The existing EMMA implementation has the drawback that road-network connections cannot be correctly drawn when multiple Focuses overlap. This paper proposes the following methods for nearby Focuses to naturally unite like water drops uniting by surface tension: 1) Focus Transformation method enables Focuses to transform in a squeezed manner to avoid overlapping. 2) Focus Union method enables overlapping Focuses to unite into a single Focus with a large Focus area. 3) Union Focus Transformation method enables a Union Focus to transform along a moving mouse pointer, enabling users to easily view map areas near the Union Focus. 4) Focus Division method enables Union Focuses to divide so that they can be operated as individual Focuses. Moreover, we have had subjects examine the experiment for testing the prototype system. Compared with the existing EMMA implementation, the proposed system required fewer operations and less time. According to the result of the questionnaires, the subjects could recognize road connections and follow the focused areas with their eyes more easily on the prototype system than on existing EMMA.

In this paper, we propose a map matching algorithm for a map information sharing system used by refugees in disaster areas. The map information sharing system stores the roads passed by a refugee as map information. When another refugee comes close to the refugee, they exchange their map information each other in ad-hoc network manner. Our map matching algorithm is based on geometric curve-to-curve matching approach and provides road segments on which a refugeemoves, from position data by GPS. In order to decrease matching errors, our algorithm adopts the point-to-curvematching method as initial matching, and incremental matching for suppressing error propagation. Experimental results shows that our proposed algorithm achieve the best result in comparison with the conventional point-to-curve and curve-to-curve map matching methods.

We propose a method for supporting presentation planning. Our objective is to support composition of scenarios according to the situation such as the background knowledge of audience and time constraints. Our idea is to introduce a presentation strategy and translate it into the policy of searching a network of content fragments in which they are related with semantic relationships. We propose a mechanism of selecting and ordering the fragments according to a strategy specified by a presenter. This mechanism allows a presenter to compose a scenario from a few fragments specified as important ones by him/her. Our method is expected to bridge the gap between the product of idea organization and presentation slides.

Collaborative filtering achieves personalized recommendation based on user collaboration. In this paper, how to preserve personal information in collaborative filtering is studied through several comparative experiments.

k

-anonymization is a standard method for guaranteeing personal privacy, in which data records are summarized so that any record is indistinguishable from at least (

k

– 1) other records. This study compares several clustering-based k-anonymization models in the context of collaborative filtering application.

Predicting traffic flow efficiently to encourage driver to avoid the sections which are going to have traffic jams is a good approach to deal with traffic congestion. Conventional prediction methods focused on specific information (e.g., speed, density, flux, and so on). However, they will consume a lot of time and storage space. This paper proposes a novel prediction approach by analyzing the aggregated information of data streams to avoid unnecessary time and storage consumption. Evaluation shows that compared with existing similar approach ES (Exponential Smoothing), this new approach can adjust its smoothing factor based on historical values and outperform in prediction results.

In this paper, we represent a novel method of spectrum enhancement for color and gray-scale images which were received under complex luminance and have dark or/and bright areas. Classical retinex algorithm permits to normalize dark areas and gives a result image with a large contract values especially for grey-scale images that does not satisfy a perceptive observer. Our Enhanced Multi-Scale Retinex (EMSR) algorithm is based on an adaptive equalization of spectral ranges not only in dark areas of image but also in bright areas simultaneously. We built a special function which stretches the spectral ranges with low and high values of intensity but reduces the spectral ranges with middle values. Also a method of image improvement after the EMSR algorithm application based on empirical dependences was designed and has shown better visual results as compared with existing filters.

Computer-assisted pronunciation training (CAPT) was introduced for language education in recent years. CAPT scores the learner’s pronunciation quality and points out wrong phonemes by using speech recognition technology. However, although the learner can thus realize that his/her speech is different from the teacher’s, the learner still cannot control the articulation organs to pronounce correctly. The learner cannot understand how to correct the wrong articulatory gestures precisely. We indicate these differences by visualizing a learner’s wrong pronunciation movements and the correct pronunciation movements with CG animation. We propose a system for generating animated pronunciation by estimating a learner’s pronunciation movements from his/her speech automatically. The proposed system maps speech to coordinate values that are needed to generate the animations by using multi-layer neural networks (MLN). We use MRI data to generate smooth animated pronunciations. Additionally, we verify whether the vocal tract area and articulatory features are suitable as characteristics of pronunciation movement through experimental evaluation

The return on experience in techniques to combat pollution show that their effectiveness depends on the situations in which they are implemented and that their choice is not trivial. From the maritime field perspective, the objective of the paper is to present a software tool to assist crisis management staff to minimise pollution impact due to maritime accidents. From a methodological perspective, the objective of the paper is to show the importance to develop ontologies (i) for structuring a domain as perceived by its actors and (ii) for building computer tools aimed to support problem solving in that domain. Such tools are endowing knowledge shared by the actors of the domain, what make them more effective within critical situations. The followed designing methodological process is based on the Cognitive Engineering method "Knowledge Oriented Design” (KOD). In the paper, the methodological process will be detailed. The resulting ontology, the architecture of the software tool and the plan generation mechanism will be presented and discussed.

Can pictures be a candidate for knowledge media? After mentioning the background of this question in chapter one, we introduce in chapter two the idea of three kinds of knowledge, from discussion of a picture from a book of Dominic McIver Lopes (Lopes 2006) —that is, knowledge about, knowledge through and knowledge in—and point out its deficiencies. In chapter three, we then propose an alternative idea of pictures from which we obtain knowledge —that is, pictures as objects and facts, pictures as process and pictures as informational indicators—and find strong support for our proposal from various research fields and practices. Finally we conclude that we could think of pictures as a candidate for knowledge media.

With the development of information and communications technology, the growth of E-learning is rapidly increased. Environment is one of the major factors for E-learning performance specially. More realistic E-leaning surrounding is affected to achievement of effective learning environment. Facial expressions are assumed to have a great impact on behavior and also on learning behavior. This attempts to transfer real user facial feature into virtual learning place. Real time facial feature detection system is developed and it is continually extracted facial expression of E-learner. The appropriate person in virtual learning environment who represents real user called ‘Avatar’ changes when the real user changes the face. In the virtual environment, the appropriate face changes are prepared to visible real user face data. In addition, any other persons able to visible user face data through web component to observe the face behavior of the E-learner.

Recently, many types of robots have been developed for not only industrial manufacturing but also interacting with human. Robots designed for human-robot interaction are expected to have the ability of communicating with human smoothly. In this paper, we propose the character giving model for

KANSEI

robot. This model makes robots individual-beings that are varied with each user. We aim to develop more humanly and empathetic robots by using this model. Robots dynamically get their own characters based on the tendency of user’s behaviors which are classified into two dimensions: dominance-submission and acceptance-rejection. Through the interaction experiments between human and the robot with the proposed model, we confirmed that proposed model could give various characters to the robot, and the character, which was given through the communication with a user, suited for each of the users.

Nowadays, it is very important to ensure the high quality of e-learning on a global scale. Therefore, the online checklist which could give a helpful guideline to ensure a high quality is required. In the current project, we chose 60 kinds of items for the checklist selected and categorized it from Japanese viewpoints of quality guarantee and established a Web check list. The system could diagnose the quality assurance by the designers themselves for e-learning materials, using self-diagnosis results. And in addition, the user can analyze his/her e-learning course statistically, according to the accumulated data on the web. We asked the teachers joining e-Help, a special group among some higher educational organizations for e-learning and credit transfer in Japan, to try the web for their self-checking. And then the data collected from the web was analyzed and the results were discussed.

This paper describes a method to visualize contrast information about two targets the Frequency and Co-occurrence Trend (FACT)-Graph. FACT-Graph is a method to visualize the changes in keyword trends and relationships between terms over two time periods. We have used FACT-Graphs as comparison method between two targets in previous research; however, the method cannot compare them as equals. To visualize contrast information, we combine two FACT-Graphs generated from different viewpoints and express the features in one graph. In case study by using 132 articles from two newspapers, we compare topics such as politics and events in them.

Social media on the Internet where millions of people share their personal experiences, can be considered as an information source that implies people’s implicit and/or explicit visual attentions. Especially, when the attentions of many people around a specific geographic location focus on a common content, we may assume that there is a certain target that attracts people’s attentions in the area. In this paper, we propose a framework that detects people’s common attention in a local area (local attraction) from a large number of geo-tagged photos, and its visualization on the “Local Attraction Map”. Based on the framework, as a first step of the research, we report the results from a user study performed on a Local Attraction Map browsing interface that showed the representative scene categories as local attractions for geographic clusters of the geo-tagged photos.

In offering intercollegiate distance lectures, problems arose from different timetables of each university in a university consortium. To address the problem, we aimed to construct a content recording and delivery system based on the distance learning system to accommodate students who could not attend the lectures. As a result of combining various systems, students were able to attend the lectures in real time or content delivery with VOD depending on the university. In addition, we suggest the possibility of encouraging students to view the lecture again.

Data extraction methods from data embedding printed images by capturing devices like scanners or mobile cameras have attracted much attention. In this technique, consideration for geometrical deformation, especially rotation and scaling, in the data extraction process is essential. This paper proposes a new correction algorithm without remarkable markers. This method exploits the log polar mapping (LPM) to detect and correct the distortion. The proposed method estimates the rotational angle and scaling factor from the amount of shift in the LPM domain. Therefore no data embedding or adding for correcting the deformation is required and the data embedded image with high quality is obtained. An image clipping technique suitable for the proposed method is also developed. From the experimental results, the effectiveness of the proposed method is shown.

Nurses are required to grasp conditions of each patient from physical, mental, social, and spiritual perspectives. From that grasp, nurses are realising both of disease control which enables observation and treatment and support of patients living. This work process is called “nursing process”. The education of assessment skill is one of the most important assignments for understanding nursing process and support system to master assessment skills waiting a long time. We developed CASYSNUPL, Computer Assisted System for Nursing Process Learning and had made use of the system in lectures and practices. It is a system that learner can imaging medical case by media files and available learning on stand-alone computer by template file used MS-Excel VBA. So learners can understand nursing process likewise learners who are learned man-to-man practice. CASYSNUPL is evaluated by learner that it helped their understand for nursing process.

Giving agents the ability to sense, recognise and appropriately respond to the human emotions is one of the main methods to make the agents more believable. In intelligent tutoring systems, learners’ affective states have been incorporated in providing adaptive feedback. In game industry, however, emotions of players have not been paid enough attention. Games normally do not take into account of players’ emotions when they play games. Non-player characters do not respond to players’ emotional status. We argue that adapting to players’ emotions can make the non-player characters more believable and the game more enjoyable.

In the current decades, compact city becomes a new concern of urban planning among most of Japanese cities. Some local governments in Japan target to realize the conception of compact city pattern through policy intervention, such as encourage households remove from suburban to downtown and then relieve the decrease of population in urban center areas. Recently one of such kind residential policy is argued by local government in Kanazawa city, Japan. This policy manages to attract and encourage households remove to downtown by offering a local housing allowance. The contribution of this work is that we developed an agent-based Household Residential Relocation Model (HRRM) for visualizing the effect of this residential policy. HRRM is built on household interaction through housing relocation choice and policy attitude, and thereby it can simulate the diversified decisions of households in all their lifecycle stages. Through the simulation of HRRM the effectiveness of this residential policy can be visualized and hereafter helps local government to view the effect of the residential policy.

Extensive reading is a method of learning language by reading many books that are easy and enjoyable. There is a need for a placement instrument to evaluate learners’ reading levels and the Edinburgh Project on Extensive Reading tests are well established. However, they have a few drawbacks, for example they are paper tests and preparing the correct level of test for individual learners is difficult. The Extensive Reading Foundation Online Self-Placement Test (ERFOSPT) is a new online self-placement test that “evaluates learners fluent reading levels easily, quickly and accurately”. It delivers the test via the Internet and adaptively selects texts depending on records of the learners reading level and the results of previous questions in the test. In this paper we explain the development of the Online Self- Placement Test engine that is a part of the ERFOSPT.

This paper presents a method (named DOSR) to support the semantic retrieval of XML documents in a specific domain. It takes the entity as the basic unit of information processing to guarantee the semantic integrity of returned results. An efficient index method named Entity-based index is designed for indexing the entities. It can greatly reduce the size of the index file while guarantee the speed of parsing entity. In order to rank the querying results, the Stratified-Weight-Method is proposed to make an improvement towards the traditional technology. Experimental results show that DOSR can infer users’ search intention effectively, locate the search target quickly and return the exact results in accordance with users’ expectation. The results processed by DOSR guarantees semantic integrity and reasonable ranking results.

Large numbers of travel traces are collected by vehicles and stored for applications such as optimizing delivery routes, predicting and avoiding traffic, and providing directions. Many of the applications preprocess the travel traces, usually composed of position data, by matching these with links in the underlying road network. This paper addresses the problem of persistent storage of large numbers of vehicle travel traces. We propose two methods for using a routing algorithm and road network to encode a travel trace formed by a sequence of links. An encoded trace, composed of a few links, is useful to store or share and can be decoded into the original travel trace. Considering that drivers tend to proceed from an origin to a destination by using the shortest path or going as straight as possible, the two proposed methods use the following two routing algorithms: a shortest path algorithm; and a following path algorithm, which finds the path that avoids turns. The experimental results for 30 real traces show that a travel trace is encoded into only 5% or 7% of its links on average using the shortest path algorithm or the following path algorithm, respectively.

Many recent studies have focused on dialogue communication. In this paper, our target is to make a robot support a communication between humans. To support a communication between humans, we believe that there are two important functions: estimating dialogue moods and behaving suitably. In this paper, we propose dialogue mood estimation model using the utterance intervals. The proposed estimation model is composed by relating the subjective evaluations for several adjectives with the utterance intervals features. Through the estimation experiments, we confirmed that the proposed system could estimate the dialogue moods with a high degree of accuracy, especially for “excitement,” “seriousness,” and “closeness.” And we suggested that the utterance intervals features had a high potential for the dialogue mood estimation.

Though vocational aptitude tests have been conducted, it sometimes judge people do not have aptitude for their desired jobs, because of weak points. If they know their weak points in early age such as elementary school ages, they may be able to overcome the points. It would extend their range of job choices. However, elementary school children cannot take most of tests because of loaf of knowledge. In this paper, we propose a method to evaluate vocational aptitudes for elementary school children quantitatively. In this method, we extract aptitude value, applying evaluation expression on operation logs. Specific evaluation expression is provided for every aptitude. We implemented aptitude test system assuming an example of the home-delivery service job experience. We have applied the system to 15 people.We have evaluated with correlation coefficient between extracted values and the result of CPS-J,Which is one of widely used vocational aptitude tests. From the result, we have found that some aptitudes should be calculated in early set while others can be derived from any set in repeated job experiences.

This study proposes the framework of a system that extracts mathematical concepts from an input mathematical expression. In this paper, math concepts are represented as specific patterns in math expressions such as “differential equations” and “quadratic functions.” This system, termed the concept extraction system, presents a math concept when an input math expression includes the pattern for that particular math concept. The system uses two key components: “MathPlaceholder,” an originally defined XML vocabulary to describe patterns, and a pattern discriminator, a mechanism to identify whether an input math expression includes the predefined pattern(s). Math expressions described by an XML vocabulary called Content MathML have been used for this study. Lastly, the following two applications of the proposed system are presented: (1) it can be used as an information retrieval tool to match math concepts in math expressions, and (2) it can be used together with a learning management system that provides study material for the concepts used in a given math expression.

In this research an e-Learning system was developed focusing on how to implement the idea of dynamic contents in teaching materials, creating what we call Dynamic Teaching Materials, with which teachers themselves can easily change or update the teaching materials according to the requirements of the class. The contents created with the system can also be easily shared between any teachers, and when the shared content of one teacher is updated all the other teaching materials that use the same content can also be automatically updated. The teaching materials are produced in a WYSIWYG interface that allows teachers to create interactive multimedia content without any knowledge of programing languages; the content can also be edited in realtime. The interface design was focused on maximum utilization of the user’s prior computer knowledge, so they do not have to learn basic operations. The system was tested and is being used by the Department of Foreign Languages and Translation of Brasilia University in Brazil.

With many remarkable advances in technology, not only studying with tutor at school but also studying through computer at home become preferable. Intelligent Tutoring System (ITS) is one of the research fields which aim to support the individual learning intellectually. To provide the learning material of the domain knowledge in many ITS, the learning materials are statically associated with each other in advance and given to student based on her/his understanding state. Motivating student and making them more interested in the learning content is the system’s task in the computer-supported systems. If students study on content which they are interested in, learning activity becomes more effective. Our research objective is to construct a system which automatically generates multiple-choice cloze exercises from text input by the student. We focus on supporting individual study of learning English grammar. In this paper, we propose a representation method of English grammar by Part-Of-Speech (POS) tags and words, the calculation procedure for estimating the understanding state of student in the student model, and the learning strategy for generating the next exercise based on the student model.

This study is an attempt to design a method for an autonomous pattern classification and recognition system for emotion recognition. The proposed system utilizes Biased ARTMAP for pattern learning and classification. The ARTMAP system is dependent on training sequence presentation to determine the effectiveness of the learning processes, as well as the strength of the biasing parameter, lambda

λ

. The optimal combination of

λ

and training sequence can be computed efficiently using a genetic permutation algorithm. The best combinations were selected to train individual ARTMAPs as voting members, and the final class predictions were determined using probabilistic ensemble voting strategy. Classification performance can be improved by implementing a reliability threshold for training data. Reliability metric for each training sample was computed from the current voter output, and unreliable training samples were excluded from the performance calculation. Individual emotional states are highly variable and are subject to evolution from personal experiences. For this reason, the above system is designed to be able to perform learning and classification in real-time to account for inter-individual and intra-individual emotional drift over time.

Honey bees correctly allocate their work force to nectar sources using the “waggle dance”. In addition, they can determine the necessity for a nectar source. Thus, they possess a value threshold for a nectar source and they can collectively regulate it. Based on the mimicry of the threshold regulation used in honey bee foraging, we are developing a system that allows agents to determine whether their own solution is worth communicating to other agents. We propose a novel bio-inspired optimization algorithm, honey bee optimization (HBO). HBO is a multi-agent system based on the foraging activities of honey bees. To test the characteristics of HBO, we applied it to the travelling salesperson problem (TSP).

Recently, 3D CG animations have become in great demand for movie and video game industries. As a result, many 3D model and motion data have been created and stored. In this situation, we need any tools that help us to efficiently retrieve required data from such a pool of 3D models and motions. The authors have already proposed a motion retrieval system using Interactive Evolutionary Computation (IEC) based on Genetic Algorithm (GA). In this paper, the authors propose a 3D model retrieval system using the IEC based on GA and clarify the usefulness of the system by showing experimental results of 3D model retrievals practically performed by several users. The results indicate that the proposed system is useful for effectively retrieving required data from a 3D model database including many data more than one thousand.

A data model incorporating the concepts of recursive graphs has been proposed for representing the contents of multimedia data. A

shape graph

, which represents the structure of a set of instances, has to catch their incremental updates. It is difficult to manage instances when they have recursive structure. This paper proposes a method of managing the recursive structure of instances. The procedure incrementally revising the information of the structure of shape graphs is presented. Owing to this procedure, the recursive structure could incrementally and properly be managed and represented in the shape graph.

Automatic collage design is a widely spread task in many applications, from a design of home collages to commercial advertising projects. We have proposed an intelligent collage system based on adaptive segmentation of Regions of Interest (ROI) in photos or video frames. Under ROI we assume images of people, their faces, animals, cars, ships, buildings or other large objects which are basically situated in the center of photos. Also we enhanced a method of a seamless blending of collage regions using flexible contours and color distribution on region boundaries. In the case of video sequence, we use an adaptive selection of frames and a special algorithm for removing of non-informative or repeatable frames. The intelligent collage system can work in automatic or handle modes, in last case with tuning of parameters. The main advantage of our approach is a great variability of different possible placements of collage regions and a high aesthetic effect of a designed collage.

In this paper, we propose an intuitive operating system for humanoid robot based on recognition and variation of human body motion. In this system, human body motion is dynamically sensed by six-axis sensors in controllers held in user’s hands. Then, the variation of human body motion is also recognized by gyro sensors in the controllers. Based on the obtained sensor information, the body motion classifier is constructed using Hidden Markov Model. By using the body motion classifier, the proposed system recognizes user’s body motion, and redundant motions of humanoid robot are prevented. User can intuitively operate humanoid robot and make it do the user’s intending motion. We conduct a task experiment to evaluate the usability of proposed system. In the experiment, for comparison with the proposed system, we prepare three methods of operation: joystick, kinect, and operating system using six-axis sensor without recognizing the variation of human body motion. As the result, comparing with other systems, we confirmed that the proposed system has more versatility and the humanoid robot could be more appropriately operated with the proposed system. And it was suggested that the proposed system enables user to operate humanoid robot, appropriately and intuitively.

A 3D urban model is an important information infrastructure that can be utilized in several fields, such as, urban planning and game industries. However, enormous time and effort have to be spent to create 3D urban models, using 3D modeling software. In this paper, we propose a GIS and CG integrated system for automatically generating 3D building models from building polygons (building footprints) on a digital map. By combining the building footprint analysis with information from GIS maps and domain specific knowledge, the complexity of the building generation process can be greatly reduced.

This paper discusses a methodology for project-based learning (PBL) for edutainment using locomotion robots. The aim of PBL is to develop a locomotion robot as a new shape of artificial creature not existing in the natural world, and to design new locomotion patterns. Therefore, the PBL is composed of two steps; (1) Students conduct conceptual design of artificial creatures, and (2) Students develop hardware design and locomotion generation based on the conceptual design. This paper introduces two examples of development of locomotion robots. Based on their experience on problems and troubles in the trial and error through the development of locomotion robots, the students learn the relationship between the shape and locomotion patterns, the tradeoff the stability and high-speed locomotion, and the difficulty of problem solving.

Processing

sum k

-Nearest Neighbor (NN) queries on remote spatial databases suffers from a large amount of communication. In this paper, we propose

RQP-M

search algorithm for efficiently searching

sum k

-NN query results to overcome the difficulty. It refines query results originally searched by

RQP-S

algorithm with subsequent

k

-NN queries, whose query points are chosen among vertices of a regular polygon inscribed in a before-searched circle. Experimental results show that

Precision

is over 0.99 for uniformly distributed data, over 0.95 for skew-distributed data, and over 0.97 for real data. Also,

NOR (Number of Requests)

ranges between 3.2 and 4.0, between 3.1 to 3.8, and between 2.9 and 3.5, respectively.

Precision

of

RQP-M

increases by 0.04-0.20 for uniformly distributed data, in comparison with that of

RQP-S

.

This paper considers the effect of social interaction on art museum behavior and cognitive/ aesthetic experience, arguing that social interaction may represent one of the potentially most detrimental elements in museum-based viewing of art—calling into question the current push to increase social interactions through museum social and knowledge media design. This is considered through three case studies with the same works of art, varying only design elements creating social interaction and considering the differences this creates. Through a psychological viewpoint, these cases are considered and social interaction’s effect is presented, considering how these findings might connect to museum and general conceptions of social and knowledge media design.

This project for learning nuclear energy safety was carried out through e-learning. Problem Based Learning (PBL) was selected as the educational tool and Metaverse as the class environment. The virtual classroom was built on a virtual island of Second Life owned by Nagaoka University of Technology. Three students from two National Technical Colleges in Japan joined the project. A teacher gave the students a short lecture and proposed the problem. Students understood the contents very well and solved the problem through chat-based discussions in Metaverse. Students’ clear and precise understanding, their high activeness of discussion and high interest for the safety of nuclear energy was apparent throughout this successful PBL class project. The results indicate very clearly that this kind of PBL class was obviously possible for actual e-learning in nuclear engineering and engineering education.

The present paper reports on online collaboration support tools for a blended and project-based learning (PBL) program. The authors have been conducting a research project in an attempt to implement e-Learning technology for project-based learning on the development of embedded software, and reported several times on each of the tools and the learning program. This paper corresponds to the final report of the research project, which clarifies the position of each tool and provides evaluation results and discussions. The tools proposed and developed include integrated repository, review support, online white-board, and postmortem support tools. A project-based learning program has also been constructed on the premise of use of those tools. A trial of the proposed program has been carried out, then a questionnaire survey and an achievement test have been conducted to evaluate the learning effect of the program and the tools. The results suggested that the program is feasible for PBL on embedded software development, and the tools facilitate collaborative activities between learners and are effective for enhancing their ability.

Recently,we can easily acquire various information frommanyWeb sites. However, it is difficult to determine the next browsing page from many Web pages. We often change the search target to narrow down the result pages found out by the existing search engine. Generally, it is not easy for a searcher to change the query before he/she knows the page content in search result perfectly. In this paper, we propose a method for specifying query terms to narrow down the pages in the search result effectively. Our method detects the representative terms of the current search target according to the currently browsed page. In order to detect the representative terms of the current search target, our method calculates the occurrences of individual terms in the currently browsed page, per constant time interval. We explain the prototype system with our method and show our experiment.

The goal of the path planning problem is to determine an optimal collision-free path between a start and a target point for a mobile robot in an environment surrounded by obstacles. In case of probabilistic environment not only static obstacles obstruct the free passage of the robot, but there are appearances of obstacles with probability. The problem is approached by the bacterial memetic algorithm. The objective is to minimize the path length and the number of turns without colliding with an obstacle. Our method is able to generate a collision-free path in probabilistic environment. The proposed algorithm is tested by simulations.

This paper proposes a method of automatic transformation of newspaper articles for generating spontaneous news speech in a news speech delivery service. Among several differences between written and spoken languages in Japanese, this study takes particular note of the importance of the differences in sentence-style, and also works on

taigen-dome

sentence, which is a Japanese-specific rhetoric style to cut off the words following a certain noun. The sentence-style transformation from written language to spoken language is a kind of personalization of a textto- speech application. We took an example-based approach to complement taigendome sentences. An experiment using Japanese newspaper articles has shown the achievement of the acceptable performance on the selection of the complementary phrase for taigen-dome sentences by using the type of the last noun of the sentence, the tense, and so on.

E-mail systems are common communication tools, and it is desirable that e-mailmessages are written readably for recipients. One of techniques to write readable e-mail messages is to insert linefeeds and blank lines appropriately. However, linefeeds and blank lines in incoming e-mails are not always inserted at positions where recipients feel readable. If linefeeds and blank lines are inserted automatically at proper positions, readability of e-mail texts are improved and recipients can read e-mail texts efficiently. This paper proposes a method for text formatting of email texts by inserting linefeeds and blank lines into incoming e-mails at positions where recipients feel readable.

In many situations, we present our ideas using presentation tools, such as PowerPoint in Microsoft Office. However, story created by author is sometimes inappropriate and listeners cannot understand topics that the author wants to insist by watching the generated slide. Our research aims at constructing the system which automatically detects differences of author’s intention and slide structure, and points out inappropriateness of generated slides. Author’s intention for each slide is captured by assigning each slide in

topic template

. On the other hand, topics that listeners may understand from generated slides are estimated based on the lexical information in the slides and are organized as

topic tree

. Relations between generated slides are detected by the change of their focusing nodes in a topic tree. By comparing relations grasped by the topic template and topic tree, the system is able to point out inappropriate slides automatically.

Problem Based Learning (PBL) is important for engineering education and has been a tool for creative engineering design. It can enhance creativity and has been used to successfully carry out many experiments in the real world. Researchers in the US and Japan (the authors) are pursuing studies to determine the effectiveness of its use in a virtual environment, one with cutting-edge technology and opportunities for complementary activities between face to face learning and electronic learning. Here students can work from anywhere in the world, at any time, and at their own pace. For this project, student teams from the US and Japan were asked to solve problems in a virtual community. Each team worked independently on a different project. The US team designed and built a car for the future, while the Japan team focused on designing a safe way for using nuclear energy. A discussion about the US team’s successful project is provided.

This paper describes a concept for a calculation exercise system used in order for students who study for SPI2 Test may exercise a numerical calculation repeatedly. We aim to develop a system characterized by generating questions dynamically to each student in order to ease the burden for the teacher when preparing many original questions. In such a case, in order to raise learning effect, it is important to measure a student’s academic ability exactly and distribute questions according to academic ability. In this paper, we propose a method to estimate students’ understanding based on an academic ability diagnostic test using item response theory, and a method to control questions to distribute using the information of hierarchical structure among questions in the study unit.

This paper proposes an automatic composition method of piano works for novices. Because, there is a shortage of piano works suitable for various learning stages in existing textbooks. At first, this study analyzed and systematized the study items in the learning process of the textbooks. This study also confirmed that the study items are adequate to represent the learning stage of each piano works in the textbooks. This study then defined the restrictions on the piano works for novices according to learning stages. Finally, this study proposes the automatic composition method based on the learning stages and their restrictions. This paper describes these learning items, their restrictions, and the proposed method. Additionally, this paper shows an example of piano works composed with the proposed method.

This paper proposes a multimodal interaction API (MMI-API) and a library for the development of web-based multimodal applications. The API and library enable us to embed synchronized multiple inputs/outputs into an application, as well as to specify concrete speech inputs/outputs and actions of dialogue agents. Because the API and the library are provided for JavaScript, which is a commonly used web-development language, they can be executed on general web browsers without having to install special add-ons. The users can therefore experience multimodal interaction simply by accessing a web site from their web browsers. In addition to presenting an outline of the API and the library, we offer a practical example of the use of the multimodal interaction system, as applied to an English pronunciation training application for Japanese students.

We considered the knowledge and skills that are required for information morals education at high schools in Japan. Clearly its knowledge and skills are critical for high school students. We demonstrate the problems with the teaching material of information morals in Japan. Teaching materials are needed to help students cope with the problems caused by utilizing the knowledge and the skills of the information society to provide a more general way of thinking that can apply other examples and a positive attitude toward the information society. We proposed teaching material based on the goal-based scenario theory. From our practice results that used teaching material whose subject was spam e-mail, we achieved the learning goals of the teaching material.

There is a growing need for a well-designed learner support system for presentation in English, particularly in non-English-speaking countries. We have developed a prototype of comprehensive learning support system of basic presentation that consists of several modules including digital contents of preliminary tutorials, an interactive aide for organizing a presentation and its corresponding slides, semi-automatic evaluation estimation, and a online review of recorded presentations. After trials, it has been clear that non-verbal aspects has to be extensively supported by such a system. In this study, we made an extensive observation and analysis of available professional and learner presentations, and extracted significant non-verbal elements in those presentations. Then, we designed learner support materials for non-verbal aspects, based on a non-verbal ontology we also designed. It is expected that the implementation of those materials will let learners to learn more effectively how to make and conduct a presentation.

Our goal is developing a system to improve knowledge co-creation skills by reflecting so-called “knowledge co-creation discussion.” The discussion is a core part of our educational program that conducts a knowledge co-creation processes that we develop in the medical service field. Meta-cognitive skill plays a key role to improve the knowledge co-creation skills especially in the field where there is no pre-defined definite answer. However, it is difficult for learners to train meta-cognitive skills since they are quite tacit, latent and context dependent: one cannot observe other one’s cognitive processes conducting in her mind as internal self-dialogue. In our research, we aim to develop learning support system by focusing on isomorphism between tacit internal self-dialogue task and discussion tasks conducted observably in the external world. In this paper, we firstly describe background of our research. Then, we describe underlying philosophy to train meta-cognitive skills essentially required for knowledge co-creation. Finally, we overview our learning support system through reflective monitoring on learners’ discussion processes.

With the advent of PowerPoint and Keynote that can effectively create attractive presentation slides, people can use them to exchange and discuss ideas together. However, because it is necessary to prepare many slides to enable audiences to understand the content, authors need to prepare the best possible slides. Our skeleton generation method is designed to help authors to prepare slides with ease by constructing slide layouts based on the expression styles that the level positions of words expressing their role in slides from the text in the textbooks they use. By analyzing the role of the words in the slides, our method can then extract the differences between the important elements in both the texts and slides. To generate skeletons for slides from target texts in a textbook, our method derives the expression styles of the words from pre-existing texts and their slides. Finally, it generates slide skeletons by using the same expression styles of the corresponding words from the target texts arranged in slides, which are the same as the layouts of pre-existing slides. We also present the results of an evaluation of the method’s effectiveness.

This paper considers the optimal facing study support interval for the learner and derive analytically the optimal interval study support policy by a stochastic model using access logging data of the e-learning system to contents. If a lecture does not provide the facing study support to the student in an e-learning system, the learner has the possibility of dropping out of the target subject to be studied. However, if the lecture indeed provides such support to the learner every time, a problem occurs from the viewpoint of its cost-effectiveness.

In this paper, we describe a practical use of social networking service to support continued communication among adult students in an e-learning program. The access counts are analyzed in functions of my homepage, friends’ diary, ranking, send message, add diary, etc. And communication activities are compared in three coures periods’ students. The results show that the most communication activity is reading friends’ diaries, the next most activities are accessing footprint and listing all diaries. Furthermore, both the first-year students and the second-year students show continued communication activities in SNS even after they have graduated. Especially, the first-year students show significant more communication activities than the other two coures periods’ students. As a result, we conclude that continued communication among adult students is supported by the SNS.

The sensitivity information processing have been developed by visualizing it through developing a method of describe, such as musical stores. We propose a knowledge media of describing tactile sense and visualize it, with refereed to the musical score. We analyze massages in a beauty salon by using tactile scores; we decompose the massage to 42 kinds of basic method components then analyze them by using the principal component analysis; found out that they are classified into 6 basic groups. We visualize the massages as transition sequences among the basic groups in the principle component space; and discover that two basic groups are intermediate one basic group to another.

The taxi is a flexible transportation system that everyone can move to any destination. However, in Japan, the charge for the taxi is more expensive than other transportation facilities. The taxi business is in a very tough situation because the cost of crude oil suddenly increased in addition to the influence of the oversupply of the taxi market. Recently, the application of Information Technologies has advanced on taxi industries (e.g., the fare payment by non-contact IC and car navigation system). One of the technologies that gain such the attention is a probe system which can store a large amount of customer trajectory data. The probe system will improve the profitability of taxi companies if the demand in the future can be forecasted from the statistics. Therefore, in this paper, we try to forecast the taxi demands from the taxi probe data by a neural network (i.e., multilayer perceptron). First, we analyze the statistics of the taxi demands and make the training data set for the neural network. Then, the back-propagation learning is applied to the neural network to reveal the relationship of regions in the Tokyo(i.e., 23-words, Mitaka-shi, and Musashino-shi). Finally, we report our discussion about the result.

In this research, the authors propose improvements to automatic lecture-archiving systems. We identify silent periods in lecture videos by analyzing the lecture’s audio content. We then create shorter videos by removing the identified silences and evaluate their encoding time and contents of those videos.

It is an important theme for pre-school teacher education to identify the merits of view to view lesson and of e-Learning method under the limitation of time or the number of instructors. In this paper,we discuss the difference and limitation of skill transfer, comparing two types of prepared learning environment. One consists of annotated scores and model performance videos served by e-Learning, while the other is supplemented with view-to-view lesson after taking the self-learning. The analysis of our experiment will indicate that e-Learning contents provide a powerful method for cognition of correct length or pitch of notes. Based on this result, we will propose a hypothesis that e-Learning is effective to learn items related to the motion control and music perception. Although students can improve their skill by the self-learning via e-Learning to some extent, they get more stimuli from view to view lesson. Those stimuli lead to more freqent awareness and the improvement of cognition of the length of notes, fine movement of notes, or tempo. In contrast, the expression of dynamics of music will turn out not to show any radical improvement in both learning methods. However, the students who take a view to view lesson can get instruction reaching at dynamics of music because of their faster mechanic skill improvement. As a result, we will find that it has some additional effect to change their attitude into the direction more oriented to improve their music expression.

We propose topic bridging as a method for story generation support. In our research, a document is defined as a story fragment and a story is defined as a sequence of story fragments. Topic bridging suggests story fragments that can function as a bridge between the start topic and the goal topic to generate a story. To do this, we propose a topic dynamics model corresponding to the story that is based on a spreading activation model, which we call the spreading topic model. On the basis of this model, we defined the term context-dependent attractiveness to indicate the dynamic popularity of a term spreading through the term relations in a new concatenated story fragment. The term context-dependent attractiveness features the topic of the story fragment. We propose the topic bridging method to estimate the feature of the story fragment that bridges the topics of the start story fragment and the goal story fragment by solving the inverse problem of the term context-dependent attractiveness.