In this paper, we discuss a method to acquire knowledge from various texts. It is better when we acquire knowledge from more than one sources that stands on various point of view than from a single one. However, it is not so easy for us to collect appropriate information of specific amount for a specified topic and find relations among them. Therefore we propose a method and a system which assists users for it. The system estimates a user's level of knowledge for a specified topic and represents appropriate texts to read.

Up to now, AR-based inorganic chemistry learning support environment using Smartphone-based HMD is proposed. The virtual environment displayed from this system is constructed from recorded image and CGs. By putting some markers in recoded area by camera of smartphone, corresponding CGs (instruments, water solutions, flame and so on) are displayed in the virtual environment. User can perform virtual experiment to learning chemical reaction relationship between ions and chemical reagents. In this paper, it is confirmed that proposed approach has some advantages to learn the knowledge about chemical reactions.

Problem-posing activities can provide a significant insights into learners' understanding about structure of problem. Finding an interesting pattern in a problem-posing learning environment is crucial to identify an important situation that learner may have difficulty to complete an assignment. This paper expects visualizations of the activity sequences to finding turning points where learners lose a way to reach the goal of an assignment. The activity sequences are considered to represent thinking process of learners and reflect their understanding and misunderstanding about the structure of problems. This paper proposes detection of ``trap-states'' that is an intermediate state of thinking in which learners have difficulty in achieving to the correct answer. As the results from an exercise detection of trap-states from real data, trap-states have found.

It is important to construct models about a domain for understanding the domain. In previous works, we developed a learning support system for learning by structuring of knowledge structure with concept map. The system required learners to construct layered concept map and provide error visualization based on the concept map. In this research, we expand the error visualization based on several concept and link which are used in Ontology engineering. In particular, this paper show various famous errors when the ontology is described and study error visualization for the errors.

At present, many discussion systems can enable users to express their opinions. On the other hand, it is difficult for the existing systems to aggregate and summarize the users' opinions. Therefore, In this study, we propose a system that allows the aggregation of their opinions. Our system actualizes the opinion aggregation by grouping users with similar opinions. The similarity of the users is calculated using collaborative filtering.

Reinforcement learning is well known with respect to acquisition of action for autonomous robot control. However, it is required much trial for its convergence. In our research, we have proposed the method of reinforcement learning addition of semi-supervised learning. The method yielded that autonomous robot learning in real environment quickly corresponds to environmental changes to goal achievement. The effecitiveness was shown in the results in the simulation. In this paper, we verify the learned action in the simulation to the real environment.

To learn the action in a dynamic environment in robotics, reinforcement learning is a method to acquire the state-action space. However, if the complex environments and complex objectives is given, the number of states increases and it exponentially caouses many attempts of learning. It takes unterminated time for convergence of the learning. In this paper, we introduce geometrical similarity of conditions, discuss the efficiency of learning. Normally, in the state-action space, eligibility of the trace is well known to improve the efficiency of learning. In order to simplify the eligibility of traces, we apply the similaritiy under scaling and rotation in the state-action space. Currently, we experimented using only the similarity of symmetry, it shows the possibilities of our method.

Problem-based learning (PBL) is one of the most important learning approaches in recent years. In PBL, participants in a group discuss collaboratively to solve a problem and to make a decision through a conversation. In this paper, we propose a tool for supporting consensus-building on multi party conversations. We call it ``Discussion Map''. It consists of nodes and links between them. We divide the nodes into two types; alternatives and criteria. Each criterion contains the importance value. Each link between nodes also contains the importance value. We evaluate the effectiveness of the discussion map system experimentally.

Various teaching strategies raise the quality of the class. Learning teaching strategies from instances of classes is useful method. However, it is difficult to learn teaching strategies from instances because instances do not describe the design rationale of themselves. This study propose a learning method to reconstruct the design rationale of instances of classes. OMNIBUS ontology can be a framework to describe the design rationale and Kit-build method can be a framework for learning by reconstruction. This paper reports the design and the development of a practice system for reconstruction of design rationale in lesson plans

This paper describes a function that generates (1) a series of fill-in-blank problems, and (2) a concept map corresponding to the problems from input of a series of propositions. As an experimental use of this function, we had used the fill-in-blank problems and the concept map (used as a kit-build concept map) in (1) science class of 5th grade in an elementary school, and (2) lectures of a university. As the results, the generated problems and concept map are able to use practically. Beside, in comparison with the fill-in-blank problems, concept map as kit-build method is obtained better learning effect than the problems.

People usually deal with much information, and judge next what to do in an instant. Therefore, it is useful to support for acquisition of an ability which can be judged instantly. In this paper, we designed a learning support system for instant judgement ability in the FPS game. The system provide a situation of its actual game that has much information, and let them judge how to behave in the situation. A learner makes himself structure tacit knowledge used for a judgement after it. This activity facilitate their reflection and to understand their error of judgement. In addition, by increasing complex of the situation gradually, our system lead deep understanding. To realize the system, we describe structuring FPS game and the judgement.