This study constructed a dataset targeting character product images, that are an important source of information about customer behavior, for business marketing. The images were collected independently and assigned detailed tagging information based on hypotheses, with the aim of training the models and contributing to multifaceted evaluation. Additionally, experiments using classification models demonstrated that the constructed dataset and tagging information contributed to verifying the characteristics of the models.

This study conducts a comparative evaluation of the YOLO and DETR models as an initial step toward the development of a cost-effective and accurate object detection system for images of character-themed products, such as stuffed animals, on social media platforms. The evaluation focuses on the detection methods and accuracy of each model, analyzing and identifying the strengths and characteristics of each model.

Currently, labeled datasets for hate speech detection in Japanese are insufficient. Therefore, we show that it is possible to construct a high-performance model using a pseudo-labeled dataset. This method can reduce the cost and time required to create a labeled dataset for Japanese hate speech detection. In the validation, we compared the performance when trained on a human-labeled dataset and a pseudo-labeled dataset. As a result, the F1 score of the human-labeled dataset was 0.5443, while that of the pseudo-labeled dataset with LLM was 0.5073. Thus, this method can be used to construct a high-performance hate speech detection model without human labeling.

Amusement parks are complex facilities with various attractions and shops, and many events for entertainment, making it a very difficult target for marketing research. Therefore, we proposed and built an opinion collection system using smartphones to efficiently collect user's opinions for improvement. In this paper, we propose and construct an opinion analysis and visualization system to support the grasping the situation of the amusement park, the planning of improvement. It is difficult for marketing researchers to extract important opinions from a large number of opinions. To cope with this problem, we propose a method to visualize emotional information of opinion on a map. The method classifies emotions for opinions based on Plutchik's wheel of emotions. The evaluation result by questionnaires is 4.75 on average out of 5 scales. The result shows that the proposed system can support the action of proposing business ideas.

A lot of people in Japan suffer from bad conditions of their mental health with an increasing tendency, in particular jobseekers and elderly persons. The classical way to solve their problem is to consult a counselor, who treats these people in a way to become aware of the core of their problem and to solve it. However, the number of well qualified counselors is limited. For this purpose, we developed a VCA (Virtual Counseling Agent) as a further evolution of a formerly developed CRECA (Context Respectful Counseling Agent). CRECA had a text interface. To much more imitate a human counselor, VCA has an image avatar and a voice conversation using the Google Cloud audio API. Further, VCA is made independent of counseling content fields. Thus, it is more generalized than CRECA or ELIZA in order for every people to easily use everywhere in every situation. Here, VCA, CRECA, and ELIZA are comparatively evaluated by questionnaire after the use of 10 college students having career problems as well as elderly persons struggling with modern IT. As a result, VCA significantly exceeded the average value of ELIZA along with the significant difference at the level of 5%. Moreover, average of the evaluation value is not worse than CRECA. Compared to CRECA, VCA does not limit the content and field of consultation. Especially, elderly persons struggling with modern IT could not use CRECA that have only text interface. It can be used generally at any time in everyday natural conversation. It can easily be used by elderly people and the digital divided due to voice conversion.

Multithreaded programs are prone to dataraces. Dataraces are known to be hard to detect and reproduce by manual effort, although they often have detrimental effects on program reliability. Automated techniques are thus demanded for detecting dataraces efficiently and precisely. There have been proposed a lot of datarace detectors so far, among which dynamic ones are promising because of their precision. However, existing dynamic race detectors incur high race-checking overheads. Even a state-of-the-art dynamic race detector, called Parallel FastTrack, fails to efficiently detect races under certain conditions, despite its attempt to parallelize race detection for efficiency. In this paper, we propose an efficient and precise parallel race detector. For our proposal, we first experimentally reveal that the load-distribution policy of Parallel FastTrack tends to skew race-checking loads to a few detection threads. We then present a simple but effective technique, called POI, for balancing race-checking loads among detection threads. POI takes race-checking loads of each detection thread into account and reduces the load skew by making each detection thread manage almost the same number of memory addresses to be checked. Experiments on several real multithreaded data-processing applications show that POI succeeded in reducing, on average, about 37% of race detection overheads, which the load-distribution policy of Parallel FastTrack would impose.

A lot of IT personnel have psychological distress and counselors to help them are lack in number. Therefore, we proposed a counseling agent (CA) called CRECA (context respectful counseling agent), which listens to clients and promotes their reflection context respectfully namely in a context preserving way. This agent is now enhanced using a body language called "unazuki" in Japanese, a kind of nodding to greatly promote dialogue, often accompanying "un-un" (meaning "exactly") of Japanese onomatopoeia. This body language significantly helps represent empathy or entire approval. Our agent is enhanced with such dialog promotion nodding robot to continue the conversation naturally or context respectfully towards clients' further reflection. To realize it, the robot nods twice at each end of dialog sentence input by clients. Here, we introduce a robot that behaves human-like by an appropriate nodding behavior. The motivation for such a more human-like robot was the extension of application fields from IT workers' counselling to people, who suffer from more social problems such as financial debt, or anxiety of victory or defeat. For such applications, it is important that the agent behaves as much as possible human-like. Here, we present an enhanced experimental evaluation. The quantitative evaluation is based on the utterance amounts of a test group of individuals. These amount with and without the nodding feature are compared. Additionally, the robots with and without nodding are compared.

The paper introduces a proposal for an automated magnetic resonance (MR) image segmentation called Case-Based Genetic Algorithm Location-Dependent Image Classification (CBGA-LDIC) and presents its evaluation results. This method finds an appropriate cell set towards efficient image segmentation. It uses location-dependent image classification (LDIC), which is integrated by genetic algorithm (GA) combined with case based reasoning (CB). LDIC is a local heuristic, which defines multiple location-dependent classifiers. Each classifier is trained by Gaussian mixture model. CBGA-LDIC decomposes the whole image into some cells, makes a set of cells, and then trains classifiers. The method is applied to knee bones, because these bone formations are similar in their location. Therefore, good combinations of cells are useful and stored in case bases. To show, that this method produces better results that other ones and to find optimal parameters, some experiments have been performed and their results are presented in this paper.

Human life can be seriously affected by unusual high solar flare. It causes serious problems such as destroying satellites, damaging electric power plants, etc. Predicting solar flare peaks is indispensable. Support Vector Machine (SVM) was used to predict the solar flare intensity based on data of the past. However, such prediction is an extremely difficult imbalanced classification problem causing a very large scale combinatorial problem. To overcome this, a local optimizer such as SVM was cooperatively or synergistically integrated with a case base and GA. This extension was called Case Based Genetic Algorithm with Local Optimizer (CBGALO). This synergetic technology outperformed around 10% over SVM only. However, GA is known for a tendency to fall into stagnation or local optima without improving. This paper presents a further extended strategy which introduces GA restarting with various operations such as mutation rate etc. especially with various but high quality cases used as the initial population at each restart. This highly synergetic technology is called CBRSGALO (Case Based integration of ReStart GA with Local Optimizer). Experimental evaluation showed that CBRSGALO outperformed around 20% over just a local optimizer such as SVM.

Nowadays, a lot of IT personnel have psychological distress. Meanwhile, counselors to help them are lack in number. To solve the problem, we proposed a counseling agent (CA) called CRECA (context respectful counseling agent). CRECA listens to clients and promotes their reflection context respectfully namely in a context preserving way. This agent can be enhanced using a body language called "unazuki" in Japanese, a kind of "nodding" to greatly promote dialogue, often accompanying "un-un" (meaning "exactly") of Japanese onomatopoeia. This body language is expected to significantly help represent empathy or entire approval. In this paper, the agent is integrated with such a "unazuki" or "dialog promotion nodding" robot to continue the conversation naturally or context respectfully towards clients' further reflection. To realize such "unazuki", the robot nods twice at each end of dialog sentence input by clients. The experimental evaluation proves such nodding is effective in counseling.

Delivery route optimization is a well-known NPcomplete problem based on the Traveling Salesman Problem (TSP) involving 20-2000 cities though human oriented factors make the problem more complex. Despite of NP-completeness, the scheduling should be solved every time within interactive response time and below expert level error or local optimality, considering human oriented factors including personal, social, and cultural factors. To cope with this, Cases and NI (Nearest Insertion) are introduced into a Genetic Algorithm (GA), based on the insight that real problems are similar to previous ones. A solution can be derived from former solutions, considering human oriented factors as follows: (1) retrieving the most similar cases, (2) modifying them by removing and adding locations by NI, and (3) further optimizing them by a GA using only NI operations. This cannot only diminish the costs to compute new solutions from scratch but also inherit many parts of previous routes to respect human factors. Experimental evaluation revealed remarkable results. Though the most effective TSP solving method LKH needed more than 3 seconds, the proposed method yielded results within 3% of the worst error rate and in less than 3 seconds. Furthermore, the proposed method is able to inherit most of the delivery routes, while LKH leads to significant changes.

To detect false information or rumors spread on Twitter on and after the Great East Japan Earthquake, a tweet credibility assessing method was proposed, based on the topic and opinion classification. The credibility is assessed by calculating the ratio of the same opinions to all opinions about a topic identified by topic models generated using Latent Dirichlet Allocation. To identify an opinion (positive or negative) about a tweet, sentiment analysis is performed using a semantic orientation dictionary. However, it is a kind of imbalanced data analysis to identify usually very few false tweets and the accuracy is a problem. The accuracy of the originally proposed method was susceptible since the sentiment opinion of most tweets was identified negative by the baseline (namely Takamura's) semantic orientation dictionary. To cope with this problem, a method for extracting sentiment orientations of words and phrases is also proposed to improve the evaluation for analyzing the credibility of tweet information. This method 1) evolutionally learns from a large amount of social data on Twitter, 2) focuses on adjective predicates, and 3) considers co-occurrences with negation expressions or multiple adjectives, between subjects and predicates, etc. The effects are proven by experiments using a large number of real tweets, in which we could detect rumor tweet much more accurately. In opposition to the baseline semantic dictionary, our method leads to succeed in imbalanced data analysis.

The paper introduces a Knowledge Acquisition and Maintenance concept for a Case Based Approximation method to solve large scale Traveling Salesman Problems in a short time (around 3 seconds) with an error rate below 3 %. This method is based on the insight, that most solutions are very similar to solutions that have been created before. Thus, in many cases a solution can be derived from former solutions by (1) selecting a most similar TSP from a library of former TSP solutions, (2) removing the locations that are not part of the current TSP and (3) adding the missing locations of the current TSP by mutation, namely Nearest Insertion (NI). This way of creating solutions by Case Based Reasoning (CBR) avoids the computational costs to create new solutions from scratch.

In this paper, we discuss a semantic and application-driven approach to estimate diversity respectively similarity in Genetic Algorithms (GA) based on a relative distance. This diversity metric can used to decide, whether or not a new individual meets a requested degree of diversity. Furthermore, the trade-off between several versions of the metric and their computational complexity is discussed. Finally, the application of this metric and a formerly developed Backtrack- and Restart GA to solve the Travelling Salesman Problem under certain real time requirements is introduced along with experimental evaluation.

Bringing sensor- and context-awareness to the Web promises to open new, tremendous business opportunities. However, context-awareness involves difficult problems such as modeling social / human affairs (e.g. law, personal preferences, etc.), handling incomplete information, and managing uncertainty of sensor data. We argue that fully-fledged, Semantic-Web-style reasoning in presence of uncertainty is neither always feasible, nor always necessary for a context-based Web. As an alternative, this paper proposes a highly expressive, simple, and efficient context representation for sensor-enabled Web services. Our representation is based on a syntax inspired by standard Semantics of Business Vocabulary and Business Rules (SBVR), enriched by annotations expressing prioritized modality and trust values. The feasibility of our approach is demonstrated via a case study for developing context/ambient-aware applications that adapt to changes in the environment.

In former work, the authors developed a modeling system for university learning processes, which aims at evaluating and refining university curricula to reach an optimum of learning success in terms of best possible best possible grade point average (GPA). This is performed by applying an Educational Data Mining (EDM) technology to former students curricula and their degree of success (GPA) and thus, uncovering golden didactic knowledge for successful education. We shifted strategy from an "eager" strategy of holding an explicit model towards a "lazy" strategy of mining with data, which is really available, holds empirically, and is not a result of "guesses" about the students' general characteristics. In particular, we utilize the educational history of the students and vocational ambitions for student modeling.

In former work, the authors developed a modeling system for university learning processes, which aims at evaluating and refining university curricula to reach an optimum of learning success in terms of a best possible grade point average (GPA). This is performed by applying an Educational Data Mining (EDM) technology to former students curricula and their degree of success (GPA) and thus, uncovering golden didactic knowledge for successful education. We used learner profiles to personalize this technology. After a short introduction to this technology, we discuss the result of a practical application and draw conclusions. In particular, we could not obtain sufficient data to establish this kind of learner profiles. Therefore, we shifted our strategy from an “eager” one of holding an explicit model towards a “lazy” strategy of mining with data, which is really available without making “guesses” what they mean (profiles). In particular, we utilize the educational history of the students and vocational ambitions for student modeling.

Due to the geographical distribution, different cognitive capacity, and different domain competency of workers or learners, many misunderstandings can occur during distributed remote collaboration, leading to inefficient discussions and undesired results. To make remote collaboration more efficient and dependable, enriching cyberspace through adaptively utilizing multimedia information is proposed and evaluated. This assesses situations of remote users through information fusion of multiple biomedical sensors and the related contexts such as user profiles. Transmitting and using such information, the system adaptively supports the distributed remote collaboration by stressing, warning, and presenting keywords/summaries in multimedia. Effects of presenting keywords/summaries adaptively depending on situations and cognitive profiles of remote members are evaluated as to the decrease of not-/misunderstanding possibilities during the explanation on the Cyberspace. The evaluation demonstrates the feasibility and usefulness of the proposed method.

A delivery route optimization system greatly improves the real time delivery efficiency. To realize such an optimization, its distribution network requires solving several tens to hundreds (max. 1500-2000) cities Traveling Salesman Problems (TSP) within interactive response time (around 3 seconds) with expert-level accuracy (below 3% level of error rate). Moreover, as for the algorithms, understandability and flexibility are necessary because field experts and field engineers can understand and adjust it to satisfy the field conditions. To meet these requirements, a Backtrack and Restart Genetic Algorithm (Br-GA) is proposed. This method combines Backtracking and GA having simple heuristics such as 2-opt and NI (Nearest Insertion) so that, in case of stagflation, GA can restarts with the state of populations going back to the state in the generation before stagflation. Including these heuristics, field experts and field engineers can easily understand the way and use it. Using the tool applying their method, they can easily create/modify the solutions or conditions interactively depending on their field needs. Experimental results proved that the method meets the above-mentioned delivery scheduling requirements more than other methods from the viewpoint of optimality as well as simplicity.

In Web-based CSCW (Computer-Supported Cooperative Work), remote members communicate their intentions in cyberspace. However, different from face-to-face communication, partners' situations including their interest, concentration, boredom, and tiredness cannot be easily transmitted. Oversight and mishearing of remote partners is often overlooked. Besides, it is further difficult to understand their real intentions sufficiently. To overcome these problems, biological sensor fusion for dependable Web-based CSCW Systems is proposed. This assesses or estimates situations of remote users through fusing information of multiple biological sensors and the related general contexts. By transmitting and using information of estimated users' situations, the system enriches the cyberspace through stressing or providing warnings by multimedia. This paper clarifies the mechanism of sensor fusion engine. In this mechanism, by making information requests to different algorithms to improve estimation accuracy based on analysis of higher layers, namely situation synthesis and analysis layers, robust face recognition is realized.

The paper deals with modeling, processing, evaluating and refining university studies. A formerly developed concept called storyboarding has been applied at a university to model the various ways to study at this university. Along with this storyboard, we developed a data mining technology to estimate success chances of curricula. Here, we discuss chances to improve these results by implementing a student profiling concept that represents the students' individual educational history and a self estimation about intellectual traits and vocational ambitions.

A modeling approach for learning processes is utilized to process, evaluate and refine them. A formerly-developed concept called storyboarding has been applied at Tokyo Denki University (TDU) to model the various curricula for students to progress in their studies. Along with this particular storyboard, we developed a data mining technology to estimate chances for success for the students following each curricular path. Here, we introduce a concept of learner profiling. The profile represents the students' individual properties, talents and preferences constructed through mining personal meta data about learning preferences.

The paper deals with modeling, processing, evaluating and refining processes with humans involved like (not only, but also e-) learning. A formerly developed concept called storyboarding has been applied at a university to model the various ways to study at this university. Along with this storyboard, we developed a data mining technology to estimate success chances of curricula. Here, we introduce a validation method for this technology and its results. Further, we discuss chances to improve these results by implementing a formerly introduced learner profiling concept that represents the students' individual properties, talents and preferences for personalized data mining.

In online business, it is important to construct sale web pages offering attractive services for popular products in order to improve access and purchase rates. However, along with these web pages, management of product DB is required to set up complicated sales contents with high efficiency and reliability. To satisfy this requirement, a sale-oriented group management method is proposed. In this method, by using a sale management web page, the automatic construction and interactive modifications of sale pages as well as the automatic/interactive update of DB for each sale product group can be done simultaneously and dynamically, synchronized with the sales.

A delivery route optimization system greatly improves the real time delivery efficiency. To realize such an optimization, its distribution network requires solving several tens to hundreds (maximum 2 thousands or so) cities Traveling Salesman Problems (TSP) within interactive response time (around 3 seconds) with expert-level accuracy (below 3% level of error rate). To meet these requirements, an Inner Random Restart Genetic Algorithm (Irr-GA) method is proposed. This method combines random restart and GA that has different types of simple heuristics such as 2-opt and NI (Nearest Insertion). Including these heuristics, field experts and field engineers can easily understand the way and use it. Using the tool applying their method, they can easily create/modify the solutions or conditions interactively depending on their field needs. Experimental results proved that the method meets the above-mentioned delivery scheduling requirements more than other methods from the viewpoint of optimality as well as simplicity.

Remote collaboration has many benefits; however, due to the geographical distribution of learners, many misunderstandings can occur during interaction. For example, learners may miss the context of the discussion or do not hear parts of it, leading to inefficient discussions. In this paper, we propose an ldquoEnriched Cyberspacerdquo approach for dependable web based computer supported collaborative learning (CSCL) to overcome these problems. The system based on the approach assesses the situations of remote users through fusing information of multiple biological sensors and the related general contexts to enrich the cyberspace. A formative scenario evaluation demonstrates the feasibility and usefulness of the approach for developing effective web-based CSCL systems.

Many misunderstandings can occur during remote interaction due to different user domain competency levels, different cognitive capacity of users as well as different user backgrounds. In this paper, we propose an adaptive keyword/summary presentation approach that aims at identifying potential misunderstandings of individual users and provide these users with effective and personalized content of the current discussion. Our approach is developed for virtual worlds and tested and implemented based on the Wonderland Project. In order to evaluate our approach, a practical scenario has been designed and tested, which demonstrates how the system enriches the cyberspace for collaboration by making adaptive use of keyword/summary presentation.

The paper is focused on modeling, processing, evaluating and refining processes with humans involved like (not only, but also e-) learning. A formerly developed concept called storyboarding has been applied at Tokyo Denki University (TDU) to model the various ways to study at this university. Along with this storyboard, we developed a Data Mining Technology to estimate success chances of curricula. Here, we introduce a learner profiling concept that represents the students' individual properties, talents and preferences personalized data mining.

Building a delivery route optimization system that improves the delivery efficiency in real time requires to solve several tens to hundreds cities traveling salesman problems (TSP) within interactive response time, with expert-level accuracy (less than 3% of errors). To meet these requirements, a multi-inner-world genetic algorithm (Miw-GA) method is developed. This method combines several types of GA's inner worlds. Each world of this method uses a different type of heuristics such as a 2-opt type mutation world and a block (nearest insertion) type mutation world. Comparison based on the results of 1000 times experiments proved the method is superior to others.

A university study is a flexible but complicated system. Therefore, many students are not able to finish their studies in the designated time. To face this problem, Tokyo Denki University introduced a Dynamic Learning Need Reflection System (DLNRS). Also, a storyboarding concept was introduced to model the network of opportunities to compose subjects towards a complete study. DLNRS supports students in scheduling a semester and storyboarding for long term career planning. Here, we introduce methods to estimate success chances for a path through a storyboard. The methods are based on AI technologies such as Data Mining and Case-Based Reasoning. By classifying the students’ given path and calculating an alternative one or a supplement, if necessary, the student gets an estimation of success chances.

Generally, learning systems suffer from a lack of an explicit and adaptable didactic design. Since E-Learning systems are digital by their very nature, their introduction rises the issue of modeling the didactic design in a way that implies the chance to apply Knowledge Engineering Techniques (like Machine Learning and Data Mining). A modeling approach called storyboarding, is outlined here. Storyboarding is setting the stage to apply Knowledge Engineering Technologies to verify and validate the didactics behind a learning process. Moreover, didactics can be refined according to revealed weaknesses and proven excellence and successful didactic patterns can be inductively inferred by analyzing the particular knowledge processing and its alleged contribution to learning success.

AI technologies for knowledge mining are commonly used in technical environments. Their application for social processes like learning processes, for example, is a quite a new challenge, which is characterized by having ldquohumans in the looprdquo. Humans' desires, preferences and decisions may be unpredictable and thus, not appropriate for modeling - at a first glance. However, in learning processes didactic variants can be anticipated and can become a subject of AI technologies. A semi-formal modeling approach called storyboarding, is outlined here. A storyboard represents various opportunities for composing a learning process according to individual circumstances, such as topical prerequisites (educational history), mental prerequisites (preferred learning styles, etc.), performance prerequisites (a requested success level in former learning activities, etc.), and personal aspects (needs, wishes, talents, aims). By storyboarding, various didactic variants can be validated by considering the average learning success associated with the different paths through a storyboard in a case study. Based on validation results, success chances can be derived for the different paths. Here, a concept and an implementation to pre-estimate success chances of intended (future) learning paths through a storyboard are introduced. They are based on a data mining technology, and construct a decision tree by analyzing former learners' paths and their degrees of success. Furthermore, this technology generates a supplement to a submitted path, which is optimal according to the success chances. This technology has been tested at a Japanese university, in which students had to compose their individual plan (subject sequences) in advance, and the technology helped them by predicting success chances and suggesting alternatives.

Learning systems suffer from a lack of an explicit and adaptable didactic design. A way to overcome such deficiencies is (semi-) formally representing the didactic design. A modeling approach, storyboarding, is outlined here. Storyboarding is setting the stage to apply knowledge engineering technologies to verify, validate the didactics behind a learning process. As a vision, didactics can be refined according to revealed weaknesses and proven excellence. Furthermore, successful didactic patterns can be inductively inferred by analyzing the particular knowledge processing and its alleged contribution to learning success.

A new concept DLNR (dynamic learning need reflection) and its system used in the education at a university are suggested. The effects, particularly on the learning of software agents, are analyzed. DLNR's goal is to increase students' learning motivation through dynamically clarifying and reflecting their learning need. To attain this, DLNR includes "prerequisite conditions", "no compulsory subjects", "payment for each subject" and "grade point average: GPA " to estimate learning results. By using a tool to realize DLNR, students design their learning need, namely their graduation timeline, by themselves to achieve their academic goal towards their job after graduation. Through taking classes, students dynamically modify the timeline reflectively according to the intermediate results shown by GPA. DLNR's effects are evaluated. Particularly, DLNR was found applicable to the learning of software agents for intelligent system assistants, through incorporating more general tools such as Storyboard

In this paper, we propose the Active Learning Method, the method to acquire the control knowledge actively by the method of trial and error. In this method, the input-output information is collected for the control object by the method of trial and error, and the controller is constructed based on the information. In the Active Learning Method, the output is decided actively and the action result is evaluated, and the data with high evaluation are modeled. This modeled pattern information becomes the behavior policy optimized based on the evaluation. For this modeling, the method called Ink Drop Spread method (IDS) is used. In this system, the object system is modeled functionally from the data by the fuzzy-like processing. By using the model of bar gymnast, the learning simulation is done for the behavior policy, and we examine the validity of this method.