Vol.2 No.1 2009

14/88

Research paper : Predictive modeling of everyday behavior from large-scale data (Y. Motomura)−11−Synthesiology - English edition Vol.2 No.1 (2009) Peripheral Technology Element CPrimary Technology Element APeripheral Technology Element BServices based onintegrated technologyGeneral Cognitive ModelPeripheral Technology Element CPrimary Technology Element APeripheral Technology Element BServices & productsby integrated technologyAuthorYoichi MotomuraThe team leader of large scale data based modeling research team, center for service research(CfSR) and the senior research scientist of digital human research center (DHRC) at National Institute of Advanced Industrial Science and Technology (AIST) in Japan. He received M.Sc from The University of Electro-Communications and then joined the Real World Computing project in the electrotechnical laboratory, AIST, MITI (the ministry of international trade and industry, Japan) in 1993. In this project, he developed Bayesian network modeling and inference system, BayoNet. His current research works are statistical learning theory, probabilistic inference algorithms on Bayesian networks, and their applications to user modeling and human behavior understanding and prediction from sensory data. He also conducts the venture company, Modellize Inc.,as the chief technology officer and applies Bayesian network technologies to many variety applications in Japan. He also wrote four textbooks on Bayesian networks, and gave many lectures on Bayesian network researches and applications. He received a best presentation award, research promotive award from Japanese Society of Artificial Intelligence,Docomo Mobile Science award and IPA super creator award. Discussion with Reviewers1 Clear description of the originality as synthesiologyQuestion and comment (Hideyuki Nakashima)The author described the difficulty of the modeling of daily-life behaviors and the significance of machine learning approaches for the modeling in chapters 1 and 2. The part should include more concrete examples for the improvement of readability of ordinary readers. The author has described “Research as a service” in chapter 2, and I believe that this part is the most essential part of the presented paper. This part should be expanded.Question and comment (Masaaki Mochimaru)I guess the paper can be understood as presenting a breakthrough-type Full Research. The core technology of the paper is the Bayesian network with sensing technology and/or interview. The result of the study is a Synthesiology that realizes the solution of problems in the real world. The method and results are not just integration of the peripheral technology. The originality of the paper as Synthesiology is the suggestion of the concept of “Service as a research”. That is a social-circulation-type Full Research. This paper has embodied the concept of Synthesiology more than the past articles in Synthesiology, so, the author should explain the concept of Synthesiology explicitly. Answer (Yoichi Motomura) I have revised the abstract, title, the concept of “Service as a research”.2 The composition of the paperQuestion and comment (Masaaki Mochimaru)This journal is not for artificial intelligence but Synthesiology, so, the author should write “the dream” realized through the integration technology in the introduction. It would be a description that the dream (the realization of daily-life-support-service by a system that understands the purpose of human behavior) and the concrete examples in order to give example images to readers. The breakthrough point for realization of the dream is “description, understanding and realization of daily-life in computer”. The difficulties of the realization of the dream are as follows: (1) human behaviors include some unclear and uncertain elements and the author has introduced the Bayesian network which is a non-deterministic modeling framework as a solution for the unclear and/or uncertain modeling, (2) the difficulty for using the Bayesian network is the necessity of a large scale dataset. This difficulty can be solved by the ubiquitous sensing technology and “Service as a research” in the real world. This composition can improve the readability of the paper.Question and comment (Hideyuki Nakashima)The logical connection between chapter 1 and chapter 2 is not good. Could you add a brief summary of the rest of the paper at the last part of chapter 1?Question and comment (Masaaki Mochimaru)The description of chapter 2 has prevented the smooth story expansion of the paper as indicated by Prof. Nakashima. I suggest three solutions as follows: (1) add the abstract to the last part of chapter 1 as say Prof, Nakashima, (2) change the position of chapter 2 and chapter 3, (3) move chapter 2 to the last of chapter 8.I recommend the solution (3), because “selection of non-deterministic approach for human behavior model”, “Bayesian network as a non-deterministic approach” and “Research as a Service” are written in the paper. For readability it is not good to show the concept suddenly. So, I suggest the following storyline.Fig. a Breakthrough model.Fig. b Society cycle model.network for recommendation and promotion, Proc. of User Modeling 2007, LNCS, 4511, 257-266, Springer, (2007). C.Ono, Y.Motomura, H.Asoh: Context-aware preference handling technologies on mobile devices, Journal of Information Processing Society of Japan, 48 (9), 989-994 (2007) (in Japanese). K.Ochiai, T.Shimokado, C.Ono, H.Asoh, Y.Motomura: Supporting movie contents marketing using Bayesian networks, Annual Conference on Japanese Society for Artificial Intelligence (2009) (in Japanese). Y.Motomura et.al.: Large scale data observation, modeling, service design and applying loop for service innovation, Journal of Japanese Society for Artificial Intelligence, 23 (6), 736-742 (2008) (in Japanse). Y.Motomura, Y.Nishida: Difficulty of behavior understanding research and a knowledge sharing framework, Annual Conference on Japanese Society for Artificial Intelligence, (2007) (in Japanese).[20][21][22][23]

※このページを正しく表示するにはFlashPlayer9以上が必要です