As for the study about the query refinement, many studies have been performed till now. In this paper, we propose a context aware query refinement system using conceptual fuzzy sets and a conceptual fuzzy sets generation method employing fuzzy clustering. In an experiment we compared the proposed query refinement system with pseudo relevance feedback, which is commonly used in information retrieval. The proposed system achieved higher precision than pseudo relevance feedback in text based image retrieval. In addition, this experiment shows possibilities to achieve better precision by combining conceptual fuzzy sets and pseudo relevance feedback.

View full abstract

The nature of the concepts regarding images in many domains are imprecise, and the interpretation of finding similar images is also ambiguous and diverse on the level of human perception. Considering these features, in this paper, images' semantic classes and the tolerance degree between them are defined systematically, and the approach of modeling tolerance relations between the semantic classes is proposed. On the basis of it, a general mechanism of representing images' semantics by associative values with predefined classes regarding a corresponding dimension is depicted. Moreover, as demonstration, the methods of generating associative values with defined classes regarding the nature vs. man-made dimension and human vs. non-human dimension are described, and experimental results of images' retrieval show the effectiveness of our proposed mechanism of representing images' semantics in improving the precision-recall performance.

View full abstract

We propose a method for affect analysis of textual input in Japanese supported with Web mining. The method is based on a pragmatic reasoning that emotional states of a speaker are conveyed by emotional expressions used in emotive utterances. It means that if an emotive expression is used in a sentence in a context described as emotive, the emotion conveyed in the text is revealed by the used emotive expression. The system ML-Ask (Emotive Elements / Expressions Analysis System) is constructed on the basis of this idea. An evaluation of the system is performed in which two evaluation methods are compared. To choose the most objective evaluation method we compare the most popular method in the field and a method proposed by us. The proposed evaluation method was shown to be more objective and revealed the strong and weak points of the system in detail. In the evaluation experiment ML-Ask reached human level in recognizing the general emotiveness of an utterance (0.83 balanced F-score) and 63% of human level in recognizing the specific types of emotions. We support the system with a Web mining technique to improve the performance of emotional state types extraction. In the Web mining technique emotive associations are extracted from the Web using co-occurrences of emotive expressions with morphemes of causality. The Web mining technique improved the performance of the emotional states types extraction to 85% of human performance.

View full abstract

In searching picture images, music, perfumes and apparels, etc., it is difficult to find the object that users want through conventional keyword search methods. To cope with this problem, the searches by kansei-words and by kansei-vectors are proposed. The kansei-vector is an array of the value that indicates a degree of each kansei-word. However, due to the gap between user's subjective kansei image value and the corresponding kansei image value stored in the database, a problem occurs that the search result is different from what users want. This paper proposes the search method to resolve such a subjective criteria gap. This method automatically decreases such gaps using the user's searching history and Fuzzy modeling. This method can avoid users' burden unlike conventional methods such as ones using questionnaires. Small size experimental results showed users can do satisfactory search based on their subject image learned quickly by the proposed method.

View full abstract

In this paper, we propose an evolution model of biological differentiation based on differences of ecological traits and reproductive isolation. We use a gene expression system n-BDD (n-output Binary Decision Diagram) proposed in order to express and optimize agent's behavior. Although the system is suitable for behavior models of agents, a conventional crossover for an n-BDD has a week point. It is needed to fix each rank of variables. Each rank of variables indicates each priority of perceptual information and it should not be fixed. This paper proposes Flexible APPLY crossover, in which the rank of variables can be changed according to each environments. Using this crossover we model a more realistic ecosystem simulation where each agents can change each rank of variables at generation shift so as to adapt themselves to the environment, we observed agents becoming to differentiate into two types. In addition, we found that the biological differentiation is caused by agents adapting strategies to their ecological traits and is accelerated by the reproductive isolation.

View full abstract

In this paper, we implement the computer aided composing system of interactive evolutionary computation with composing model. The composing model is a kind of probabilistic deterministic finite-state automaton, which learns melodies and composes the melody based on learned data as an individual in interactive evolutionary computation. The composing model learns the favor model of the user and compose the user's pleasant melody. With GUI, the user evaluates and improves the melodies that composing models compose, and composing models learn the melodies that are selected from them. We suppose that it is possible that the composing model is optimized into the favor model by learning melodies of the best selection. We decided that the composing models in this system learned 7 melodies, since the results of experiment showed that the composing models learned 7 melodies composed the good melodies. In the experiment of executing the system, average fitness of the composing models was increasing according to the generation number and the average fitness stopped increasing at 12 generation, and all generated melodies are almost same at 12 generation. The offspring of composing model with high fitness had high fitness, because the melody of model of high fitness is similar with one of its offspring. The evaluation of the same melodies by a user was not constant, since the user changed the favor through the iterative process of improving or listening to the various melodies. We suppose that the system does not only learn the user's favor with composing model but also makes user's favor clear.

View full abstract

This paper proposes the clustering procedure in group decision-making environments. We develop a fuzzy classification matrix according to the evaluation vectors gathered from each decision maker. Members of the group are clustered based on the nature of the classification matrix in which the presence of transitive law is verified. Without similarity relation, the power of the proposed matrix is introduced as a negotiation value among different ideas since it always holds the transitive nature. Then the group evaluation is finally obtained. Fuzzy classification matrix enables us to organize the evaluations of all the members at a time and to assess the similarity of participants effectively. Further, consistent clustering is mathematically realized due to the introduction of transitive law. This approach allows decision makers to conduct a reasonable group decision-making in the context of different objectives while integrating or adjusting diverse views and ideas.

View full abstract

We examine effects of the generation method of scale-free networks with a small clustering coefficient and their power-law exponents on the synchronization of coupled oscillator networks. A modified Kuramoto model is introduced as a model of coupled oscillator networks. Networks employed for examinations are generated by a preferential attachment and configuration model. Under these conditions, the relationship between the coherence R and the power-law exponent γ_T is investigated by numerical computation. In addition, the relationship between the R and the coupling strength σ is also investigated. As a result, it is found that (1) the global phase synchronization occurs in the large number σ independent of network generation methods and γ_T, (2) R gradually decreases as γ_T increases under the global phase synchronization, (3) the coherence R depends on network topologies in the range of 2 < σ < 7, where γ_T, which strongly affects the coherence too, increases as σ increases, and (4) the relaxation time of coherence depends on network topologies as well as R in the range 2 < σ < 7.

View full abstract

The authors are suggesting a methodology for an estimation method of learning strategies. It involves learning sequences based on instructional structural graphs for learners and others, and what kind of learning strategies these are based on. This methodology makes use of parameters that represent learning strategies. However, in the estimation method, even though the parameters for the learning strategy introduced a Fuzzy theory concept, when the estimation was conducted, the membership function of the learning strategy was not treated in a Fuzzy manner. In addition, the estimation was carried out under the assumption that the learning sequence was constructed on a single strategy. According to this theory, when the learning strategy is estimated, the concept of Fuzzy decision-making is introduced, and the Fuzzy estimation theory for learner strategy is newly suggested. We also propose the estimation method for learning sequences that have been judged to come from multiple strategies. The objective of this research is to introduce this theory as one part of acquiring lesson planning skills for newly-appointed and student teachers.

View full abstract