The Discovery Science Project in Japan

Abstract

The Discovery Science project in Japan in which more than sixty scientists participated was a three-year project sponsored by Grant-in-Aid for Scientific Research on Priority Area from the Ministry of Education, Culture, Sports, Science and Technology (MEXT) of Japan. This project mainly aimed to (1) develop new methods for knowledge discovery, (2) install network environments for knowledge discovery, and (3) establish Discovery Science as a new area of Computer Science / Artificial Intelligence Study. In order to attain these aims we set up five groups for studying the following research areas: (A) Logic for/of Knowledge Discovery (B) Knowledge Discovery by Inference/Reasoning (C) Knowledge Discovery Based on Computational Learning Theory (D) Knowledge Discovery in Huge Database and Data Mining (E) Knowledge Discovery in Network EnvironmentsThese research areas and related topics can be regarded as a preliminary definition of Discovery Science by enumeration. Thus Discovery Science ranges over philosophy, logic, reasoning, computational learning and system developments. In addition to these five research groups we organized a steering group for planning, adjustment and evaluation of the project. The steering group, chaired by the principal investigator of the project, consists of leaders of the five research groups and their subgroups as well as advisors from the outside of the project. We invited three scientists to consider the Discovery Science overlooking the above five research areas from viewpoints of knowledge science, natural language processing, and image processing, respectively.The group A studied discovery from a very broad perspective, taking into account of historical and social aspects of discovery, and computational and logical aspects of discovery. The group B focused on the role of inference / reasoning in knowledge discovery, and obtained many results on both theory and practice on statistical abduction, inductive logic programming and inductive inference. The group C aimed to propose and develop computational models and methodologies for knowledge discovery mainly based on computational learning theory. This group obtained some deep theoretical results on boosting of learning algorithms and the minimax strategy for Gaussian density estimation, and also methodologies specialized to concrete problems such as algorithm for finding best subsequence patterns, biological sequence compression algorithm, text categorization, and MDL-based compression. The group D aimed to create computational strategy for speeding up the discovery process in total. For this purpose.KeywordsKnowledge DiscoveryNatural Language ProcessingNetwork EnvironmentInductive InferenceInductive Logic ProgrammingThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.