Yukihiro Itoh

Interactive image retrieval by natural language

This paper presents a method of building an image data retrieval system that can accept Japanese sentences and handle subjective expressions. A naive user who has little knowledge about objects in the database is likely to use subjective words to explain what he or she wants, for instance, ‘‘show me a cute one,’’ or ‘‘I would like to have a simpler one.’’ Objective interpretation of those expressions is difficult but is indispensable to retrieval systems. In this paper we propose a technique for matching subjective expressions with color features and discuss usability of our natural language interface.

Code-reading support environment visualizing three fields and educational practice to understand nested loops

In this paper, we describe a code-reading support environment and practical classroom applications using this environment to understand nested loops. Previously, we developed a code-reading support system based on visualization of the relationships among the program code, target domain world, and operations. We implemented the proposed system in exercises with nested loops. The evaluation results suggested that students could frequently fulfill learning objectives using the proposed system. However, we also discovered that some students experienced a learning impasse in the classroom. We attempted to address these students with two supporting approaches: bridging the gap between the generalization structures in the program code and their corresponding operations and enabling learners to predict the behavior of the nested loops. In this paper, we extend our previous system with new functions based on our two supporting approaches. Further, we implement the system in another classroom for nested loops. We describe a correlation between the proposed system and an understanding of nested loops using pre-/post-test comparisons. We discuss how code reading using the proposed system allows learners to cultivate a superior understanding of the program code.

Three Phase Self-Reviewing System

This paper introduces an electronic report submission system that helps effective learning of algorithms and programming. It proposes a three-phase reviewing system that involves self-reviewing of algorithms, self-reviewing of programs and staff reviewing. This is an improvement of our existing two-phase reviewing system that only supports the latter two phases. In the additional phase for algorithmic checking, learners describe an algorithm graphically using PAD, compile it, and execute it to verify their algorithm first without being troubled by syntax of a programming language; this supplies effectiveness to the efficient self-reviewing system.

Text understanding using a two-dimensional spatial imagery

This paper describes the framework of a system that can understand a textbook for machine operation, based on the model of an object world that can retain informations not only by symbolic representation but also by image. In any understanding of the description for machine operation, spatial information is important, and there must be a model that can retain the informations with natural reflection of the continuity and relative location, independently of the particular point of observation. Such a representation is difficult, dealing only with the symbolic knowledge representation used in the conventional language understanding system. From such a viewpoint, this paper notes the usefulness of the representation by the image and proposes the imagery object world model, which can utilize partially the properties of the image, using mathematical expressions in a two-dimensional real coordinate system. Mechanisms both to execute the simulation on the proposed model and to read the text using the simulation mechanism, are investigated. A system is constructed experimentally based on the proposed idea, which can construct the model for the input sentence, while executing a simulation and generating symbolic recognition expressions. With those functions, the identity or the similarity of the language expressions derived from the same object or event from different viewpoints, can adequately be decided.

Practices of algorithm education based on discovery learning using a program visualization system

In this paper, we describe three practical exercises relating to algorithm education. The exercises are based on a learning support system that offers visualization of program behavior. Systems with the ability to visualize program behavior are effective to promote the understanding of algorithm behavior. The introduction of these systems into an algorithm course is expected to allow learners to cultivate a more thorough understanding. However, almost all existing systems cannot incorporate the teacher’s intent of instruction that may be necessary to accommodate learners with different abilities by using a different instructional approach. Based on these considerations, we conducted classroom practice sessions as part of an algorithm course by incorporating the visualization system we developed in our previous work. Our system visualizes the target domain world according to the visualization policy defined by the teacher. Our aim with the practical classes is to enable learners to understand the properties of algorithms, such as the number of comparisons and data exchanges. The contents of the course are structured such that the properties of an algorithm can be understood by discovery learning in the practical work. In this paper, we provide an overview of our educational practices and learners’ responses and show that the framework we use in our practices can be established in algorithm classes. Furthermore, we summarize the requirements for the inclusion of discovery learning in the algorithm classes as the knowledge obtained from our practices.

Monitoring system for the effective instruction based on the semi-automatic evaluation of programs during programming classroom lectures

In this study, we developed a programming practice monitoring system to facilitate teachers to give appropriate instructions to students at the appropriate time during classroom lectures. To help teachers to provide appropriate instruction to learners, we identified parameters that would be useful for teachers during programming exercise in classroom lecture. We constructed a monitoring system with five functions. The system automatically acquired the programs written by students to evaluate their performance, and the teacher can obtain their performance using the five functions. We asked four subjects to test our proposed monitoring system during a simulation of a classroom lecture. The evaluation revealed that the system had a high accuracy in evaluating student programs.

Word Sense Disambiguation Using Pairwise Alignment

In this paper, we proposed a new supervised word sense disambiguation (WSD) method based on a pairwise alignment technique, which is used generally to measure a similarity between DNA sequences. The new method obtained 2.8%-14.2% improvements of the accuracy in our experiment for WSD.

The method to visualize the domain-oriented-explanation of program's behaviors

In this paper, we discuss the method of visualization of the explanations generated by our educational system. The system gives domain-oriented-explanations of programs. Our system can help learners to understand model programs and to find bugs in their programs by themselves. It generates explanations of the programs behaviors. Our previous system can output verbal explanations. However it is not easy for novice learners to get a concrete image of behavior of the program. Therefore we are trying to add a facility of generating explanations by visualizing the state of the world.

Classroom practice for understanding pointers using learning support system for visualizing memory image and target domain world

Pointers are difficult learning targets for novice learners of C programming. For such difficult targets, introducing a system visualizing program behaviors is generally expected to support learners to understand the targets. However, visualization in existing systems often conceals the concrete value of variables such as pointers; the way in which each visualized object is located on the memory is not made explicit. In order to address this issue, we focused on a program visualization system called TEDViT. It visualizes simultaneously and synchronously the memory image that is the field that presents the concrete value of variables and the target domain world that is the field that presents logically the data structures processed by the program. We consider that observing and comparing program code, memory image, and target domain world with TEDViT could work for understanding pointers. TEDViT visualizes the status of the target domain world according to the visualization policy defined by the teacher in order to allow teachers to set their instruction content based on the growing variety of learner background knowledge. We also consider that this feature could support teachers’ instructions and class managements appropriately, and improving teachers’ performance by TEDViT’s support would bring improvement of learners’ understanding. We conducted classroom practice for understanding pointers in connection with a memory model, thus introducing TEDViT to a real class. Analysis of answered scores in a questionnaire conducted after the practice suggests that our practice using TEDViT provided useful supports for participants to understand pointers. It also suggests our practice had a certain effect to reduce uneven levels of understanding among participants. Based on these results, we describe that classroom practices in our framework could support learners to understand pointers and support teachers to manage the class.

Practice supporting system with related problem set generator based on targeted educational effects

Exercises with well-designed similar problem sets are effective in classrooms. In this case, teachers design similar problem sets related to the educational effects they have targeted. However, to design these “related problem sets (RPSs)” is not so easy for teachers, especially for students who are studying the problems. To support them, an intelligent tutoring system is expected to generate RPSs for teachers’ and learners’ targeting educational effects and support exercises for learners using these RPSs. It is useful for teachers who provide RPSs to learners with their educational effects and/or learners who want to study by themselves to get rid of their own weakness. This paper suggested the RPS generation and exercises supporting functions by an intelligent tutoring system for high school chemistry named Intelligent Practice Supporting System (IPSS). Some experiments confirmed that the performance of RPS generation and the exercises with IPSS had better educational effects than the ones without RPSs.