Youzou Miyadera

A data structure for triangular dissection of multi-resolution images

In this work, the heterogeneous rectangular dissections that represent multi-resolution images of raster data are considered. Specifically, heterogeneous rectangular dissections are changed to triangular dissections in order to provide more effective feature extraction. We propose a method of generating triangular dissections that maintains “octgrid” properties and have developed a list structure suitable for extracting image features (ridges, valleys, etc.) from terrain maps. We propose a detailed list structure called “H12Code” and present examples of feature extraction using H12Code lists.

Methods for extracting the polish contexts in research document creation

Intellectual document creation and its polish represent an important phase in research activities. This kind of documenting consists of multiple tasks. Such combined documenting involves the contexts in their processes of polish. Polish contexts are often valuable hints that can be utilized for later work. However, it is difficult for most users, especially research beginners, to manage the polish contexts and satisfactorily utilize them. This research was aimed at developing methods for extracting the contexts and organizing them so that users could easily reflect on the contexts. A method for extracting important portions of the generated context information network is also proposed. The effectiveness and features of the proposed methods are discussed based on the results obtained from experiments.

Sequential pattern mining method for analysis of programming learning history based on the learning process

This research aims to realize a novel method for learning history analysis based on the learning processes in programming exercise classes. This paper proposes the sequential pattern mining method specialized for analysis of learning histories of programing learning. This paper initially describes a data processing method which investigates learning transitions as sequences based on the analyses of learners source codes and compile errors generated in their exercises. Next, this paper describes an analysis support tool. This tool assists collection of learning histories, generation of sequence based on analysis of the histories, extraction of the noteworthy patterns based on SPADE algorithm and acquisition of findings from the extracted patterns. This tool enables to effectively analyze the relationships between learning processes in programming exercises and learning situations. Such analysis can contribute to practical grasping of learning situations in accordance with learning process and acquisition of advanced findings based on it.

Methods for strategic accumulation of context information in research activities

Research activities normally consist of numerous works such as paper composition, discussions and surveys of literature. It is important for smooth accomplishment of research to strategically accumulate the context information (i.e. processes of the work, methods and results) and to skillfully utilize them. However, it is difficult to accumulate relevant context information since most automatic methods target only the specific work and/or researchers tend to selectively use different applications suitable for each purpose. This research aims to develop methods of organizing context information across the different styles of information and applications. This paper outlines methods to strategically accumulate context information and to organize it. Moreover, it describes an experiment using test-case data and discusses the effectiveness of the methods in light of the results.

Extraction methods of e-mail discussion processes considering diversity of description granularity and their complicated relationships

Discussion using e-mail is frequently conducted in intelligent works such as research activity. It is important but generally difficult to successfully grasp the processes of the discussions and their results. There exist descriptions of diverse particle sizes in actual discussion using e-mail. This feature brings about the difficulty in grasping the discussion processes in addition to basic features such as increase of number of e-mails. Although there have been some research projects which aim at extraction of discussion processes, these existing methods have not consider the diversity of description granularity. This research has aimed to develop a support system for extracting discussion processes and their visualization. This paper mainly describes methods for extracting the discussion processes considering the diversity of description granularity and complicated connections between them. This paper also describes the overview of target service and the outline of support system.

Learning History Transition Graphs for understanding the programming learning situations

This research project aims at enabling easier understanding of a learning situation on the basis of a students learning process (compile error history) in programming exercise currently conducted at universities. We express transition of learning history of a student in graph form of time series (we call it Learning History Transition Graph (LHTG)). Facilitators or a learner can refer to an LHTG along with additional information regarding the learner. Moreover, this paper describes a prototype of the support system for grasping the learning situation introducing LHTG along with the flow of our support. Application of this system enables a novel learning support with an emphasis on context.

Gaze network extraction from bookmarks in accordance with search intentions

Opportunities for Web exploration have been remarkably increased. In particular, complicated Web search with a stack of search and acquisitions of Web pages is significant in long-term intellectual activity. In such exploration, it is important to identify the noteworthy pages and their relationships in increasing bookmark in accordance with change of search intentions. However, it is difficult to grasp them since bookmarks tends to increase according to various search intentions and their changes. In this research, we develop a method for extracting the relationships between Web page by several different related elements from bookmarks. We also develop a method for extracting the gaze network based on centrality analysis. This paper mainly describes the framework of our proposed methods.

Sequential Pattern Mining System for Analysis of Programming Learning History

The overall goal of this research is to establish the methodology for analyzing learning history data of programming exercise in accordance with learning processes. To achieve this goal, we developed a theoretical method of sequential pattern mining specialized for learning histories in programming exercise. On the basis of this method, we designed a system for analyzing the programming learning history data. This system consists of functions that are responsible for collection of learning histories, generation of sequence from the collected learning histories, extraction of noteworthy patterns from a set of sequences, and acquisition of findings from the extracted patterns. This paper mainly describes the functions of the system and their implementation along with an overview of the sequential pattern mining method.

Strategic Accumulation and Organization of the Polish Contexts in Research Document Creation

Abstract Documentation in research activities consist of multiple works. Such combined documenting involves the contexts in its processes of polish. Polish contexts are often valuable hints that can be utilized for later works. However, it is difficult for most users, especially beginners, to manage the polish contexts and utilize them satisfactorily. This research aims to develop methods for organizing the contexts along with their strategic accumulation. A method to extract the important portion of the accumulated context information network is also proposed.