This paper proposes an interface for supporting a user's activity of mining useful information from a large amount of text. Finding useful knowledge from texts requires a trial-and-error process; thus, a user’s information request becomes clearer with the execution of the process. Our proposed interface attempts to expedite this process. When a user performs a text mining task, it is necessary for him or her to use combinations of different text analysis tools to analyze the text. In order to combine different text mining tools, the Total Environment for Text Data Mining (TETDM) was recently proposed. However, currently, the interface of TETDM is not sufficiently good to perform such a task; it does not provide a means to combine/switch text analysis tools smoothly. In this study, we redesigned the interface of TETDM such that it enables smooth switching/combining of text analysis tools. The proposed interface includes a graph in which nodes denote the tools and links denote the processing flow between them. The interface facilitates (1) combining/switching text analysis tools by directly manipulating nodes, and (2) understanding the current state of each tool.

Web forums become the means of online communication and information sharing sources for the learning about healthcare and related treatment knowledge. By adopting web crawlers and natural language processing techniques, the automatic identification and classification approach of the concerned HIV-related messages is proposed to facilitate the health authorities or social support groups in instant counseling. The proposed supervised GA/k-means for classification approach can help effectively distinguish “concerned” and “unconcerned” HIV-related messages on the web forum.

This paper proposes the user modeling method that reflects user's personal values for recommender systems. Existing methods such as collaborative and content-based approach tend to be less-accurate for new users and items owing to the lack of the relation between items and users' preference. Meanwhile, personal values have been taken notice because of its significant relation to potential preferences of users. While existing recommender systems usually employ user preference of items to make recommendations, proposed method focuses on users' personal values, which mean value judgments that show what attributes users put a high priority. By analyzing the relation between ratings for an item and attributes, user's priority on each attribute is extracted as a user model that reflects user's value judgment. The proposed method is applied to customer reviews on Kakaku.com, of which results show that the attributes on which users put high priority can be modeled with less reputation information.

In this paper, we used ACSM Metabolic equations and physical activities recognition technology to estimate phsical activities amount and calorie consumption based on what kind of physical activity people do. Also we describe and evaluate a system that uses phone-based accelerometer and gyroscope to perform activity recognition. The physical activities we tried to detect were walking,running,bicycling,stationary,going upstairs,going downstairs while they carry their mobile phone in the front trousers pocket .These activities are the most common for the people living in cities everyday. We explore orientation-independent features extracted from several components in acceleration. Our approach achieves over 95% accuracy in 5 cross validation for six physical activity. For going upstairs and going downstairs,which is the hardest to be recognize,the accuracy is over 85%.

In this paper, we are concerned with a problem of finding contrast concepts. If someone is interested in studying the cultural differences between different places for a common topic with much attention according to local news reports, it is usually necessary for him/her to read all news at those places and compare them one by one. Obviously, the amount of news is huge and it is extremely difficult and time-consuming. In order to help this kind of task, we try to extract contrast concepts which represent differences between two databases by using Formal Concept Analysis (FCA). To reduce the complexity of constructing concept lattices, we compress concept lattices by using Spectral clustering. We design a top-down depth-first search algorithm equipped with branch-and-bond pruning techniques for mining contrast concepts that drastically reduce the computational complexity.