人工知能学会論文誌
Online ISSN : 1346-8030
Print ISSN : 1346-0714
検索
OR
閲覧
検索
21 巻 , 2 号
選択された号の論文の9件中1~9を表示しています
    • |<
    • <
    • 1
    • >
    • >|
一般論文
論文
  • 徳久 良子, 寺嶌 立太
    21 巻 (2006) 2 号 p. 133-142
    公開日: 2006/01/11
    ジャーナル フリー
    The performance of a non-task-oriented conversational dialogue system greatly depends on whether it can generate high involvement during conversation with users. In this paper, we clarify the types of utterances concerning involvement in human-human conversational dialogue. First, we define Dialogue Acts(DAs) and Rhetorical Relations(RRs), and propose a method for measuring ``Involvement''. Next, we show that the inter-annotator agreement on these tag schemes is quite high. Finally, we investigate the relationship between DAs/RRs and Involvement. We found that affective utterances and cooperative utterances are significant to generate high involvement in conversational dialogue.
    抄録全体を表示
  • 齋藤 悠, 井上 克巳
    21 巻 (2006) 2 号 p. 143-152
    公開日: 2006/01/11
    ジャーナル フリー
    We investigate induction from the viewpoint of nonmonotonic reasoning. Induction we consider in this paper is descriptive induction. Hypotheses from descriptive induction have the weak property that they only describe rules with respect to the observations and do not realize an inductive leap. In this paper, we define a new form of descriptive induction with circumscription and the idea of explanation and show two procedures for computing it. The new descriptive induction is called circumscriptive induction. By deciding the roles of predicates in circumscription, we can intentionally minimize models of a given inductive problem. By adopting the idea of explanation, we can distinguish between background knowledge and observations. Additionally, we consider the relationship between the way of choosing the roles of predicates in computing circumscription and the property of hypotheses obtained by circumscriptive induction. It is shown that hypotheses from circumscriptive induction reflect a difference between background knowledge and observations and do not realize an inductive leap. We also investigate revision of hypotheses which is as important as generation of hypotheses. In a process of hypothesis revision, a difference between previous induction and circumscriptive induction is clearly characterised.
    抄録全体を表示
  • 今福 啓
    21 巻 (2006) 2 号 p. 153-166
    公開日: 2006/01/24
    ジャーナル フリー
    In this paper, several agents construct the multi-agent system, and each agent defines its action to maximize the reward that can obtain from the environment without communicating each other. However, if the environment around the multi-agent system changes, the action which can obtain the high reward also changes, so each agent should adapt to the change of the environment to obtain the high reward. Therefore, each agent is required to recognize the change of the environment through the acquired reward and should learn which action will obtain a high reward.
    To adapt to the change of the environment, we propose a new learning method for multi-agent system. In the proposed method, each agent has a matrix named ``transition probability matrix'' that expresses which action will obtain the high reward in the future time. Each agent updates the element of the matrix by using not only the acquired reward but also the entropy of the matrix. The update procedure of the matrix is classified into three cases according to the increase or the decrease of the acquired reward and the entropy of the matrix in the past time.
    Some simulations were done by using the proposed method. The results show that each agent can adapt to the change of the reward and obtain the high reward.
    抄録全体を表示
  • 松本 光崇
    21 巻 (2006) 2 号 p. 167-175
    公開日: 2006/01/24
    ジャーナル フリー
    In this paper, I discuss the problems of ``order in social situations'' using a computer simulation of iterated N-person prisoners' dilemma game. It has been claimed that, in the case of the 2-person prisoners' dilemma, repetition of games and the reciprocal use of the ``tit-for-tat'' strategy promote the possibility of cooperation. However, in cases of N-person prisoners' dilemma where N is greater than 2, the logic does not work effectively. The most essential problem is so called ``sanctioning problems''. In this paper, firstly, I discuss the ``sanctioning problems'' which were introduced by Axelrod and Keohane in 1986. Based on the model formalized by Axelrod, I propose a new model, in which I added a mechanism of players' payoff changes in the Axelrod's model. I call this mechanism norm-internalization and call our model ``norm-internalization game''. Second, by using the model, I investigated the relationship between agents' norm-internalization (payoff-alternation) and the possibilities of cooperation. The results of computer simulation indicated that unequal distribution of cooperating norm and uniform distribution of sanctioning norm are more effective in establishing cooperation. I discuss the mathematical features and the implications of the results on social science.
    抄録全体を表示
  • 長谷川 禎彦, 伊庭 斉志
    21 巻 (2006) 2 号 p. 176-183
    公開日: 2006/01/24
    ジャーナル フリー
    Artificial Immune System has been regarded an effective powerful optimization framework because of its powerful information processing capabilities. Natural immune system has many features such as memorizing ability, singularity against antigens, flexibility against dynamically changing environments, and diversity of antibody. Up to now, several algorithms inspired by these immune features have been proposed and applied to many problems. However, Genetic Programming with immune features which is capable of solving multimodal problems has not been proposed. This paper proposes an optimization algorithm named Multimodal Search Genetic Programming (MSGP), which extends GP by introducing the immunological feature so as to solve the problems with multimodal fitness landscape. We empirically show the effectiveness of our approach by applying the algorithm to the gene classification problem and the HP protein folding problem.
    抄録全体を表示
  • 武内 雅宇, 林 雄介, 池田 満, 溝口 理一郎
    21 巻 (2006) 2 号 p. 184-194
    公開日: 2006/01/24
    ジャーナル フリー
    As organizations evolve, it is not only important to design a collaborative workplace in which important knowledge, skill and competence are created and inherited, but also make sure that the necessary capabilities for creating and inheriting these exist. The design of such a workplace can be a series of challenges. One of the reasons is that practical and educational viewpoints must come together. The aim of this research is to conceptualize and support the design of a collaborative workplace that blends both these viewpoints within a computerized support system. To achieve this goal, we propose a collaborative workplace ontology, which we then use to describe collaborative workplace patterns. A conceptualization of collaborative workplace ontology was made based on group formation and interaction in terms of knowledge management theory and learning theory, as well as in terms of the implicit assumptions the theories appear to make in these regards. The collaborative workplace patterns, based on the preceding conceptualization, while blending practical and educational considerations, also enabled to distinguish between the two. This research is consequently concerned with the development of a support system for the design and development of a collaborative workplace.
    抄録全体を表示
  • 永田 裕一
    21 巻 (2006) 2 号 p. 195-204
    公開日: 2006/01/31
    ジャーナル フリー
    The edge assembly crossover (EAX) is considered the best available crossover for traveling salesman problems (TSPs). In this paper, a modified EAX algorithm is proposed. The key idea is to maintain population diversity by eliminating any exchanges of edges by the crossover that does not contribute to an improved evaluation value. For this, a new evaluation function is designed considering local diversity loss of the population. The proposed method is applied to several benchmark instances with up to 4461 cities. Experimental results show that the proposed method works better than other genetic algorithms using other improvements of the EAX. The proposed method can reach optimal solutions for most benchmark instances with up to 2392 cities with probabilities higher than 90%. For an instance called fnl4461, this method can reach an optimal solution with probability 60% when the population size is set to 300 -- an extremely small population compared to that needed in previous studies.
    抄録全体を表示
論文特集:近未来チャレンジ
論文
  • 古川 康一, 清水 聡史, 吉永 早織
    21 巻 (2006) 2 号 p. 205-214
    公開日: 2006/02/03
    ジャーナル フリー
    In this paper, we propose a framework for representing performance skill. Firstly, we notice the importance of performance skill representation. We introduce five different representation targets: performance tasks, performance rules, pre-shaping actions, dynamic integrity constraints, and performance states. Performance task description consists of a sequence of performance tasks and expressions. It acts as a goal description in planning. Performance rules describe model performance methods for given tasks including how to shape body parts and how to use various muscles. Pre-shaping action rules are similar to performance rules. Its role is to pre-shape in between consecutive tasks to prepare for the next task. Dynamic integrity constraints specify constraints to be satisfied during performance. They provide such general rules as prohibiting simultaneous strong activations of agonist and antagonist. Performance states are for describing real performance done by players including professionals and amateurs. The aim of the framework is to provide a uniform scheme for representing model performance methods given performance score such as music score. The representation framework will define targets of inducing formal skill rules as well as describing performance states automatically from biomechanical performance data. It also is related to a fundamental research issue of attributes finding/selection in discovering useful rules for skillful performance. We conclude our paper by stating future research direction.
    抄録全体を表示
ショートノート
  • 平田 圭二, 片寄 晴弘, 笠尾 敦司, 宮田 一乘, 原田 利宣
    21 巻 (2006) 2 号 p. 215-218
    公開日: 2006/02/03
    ジャーナル フリー
    We held the third meeting at the 19th Annual Conference of JSAI in the series of near-future challenge of Establishing Case-Based Design Support and Assessment Framework. In this Research Notes, following the presentations and discussions made at the meeting, we investigate how we bridge the gap between our design support technology and the contributions to public. We first discuss design support technology for contributing to society development in general, and, then, apply a brief case study to a picture generation system, regarding community support as a contribution to society development. Finally, we briefly mention future work and present ``call for participants'' for our next session in 2006.
    抄録全体を表示
    • |<
    • <
    • 1
    • >
    • >|
feedback
Top