Abstract:
Uncertainty, an inevitable problem in problem solving can be reduced by seeking the advice of an expert in terms of computer algorithms such as machine learning. Machine learning encompasses different types of solutions. In the present investigation, a different problem-solving scenario called game playing is investigated. For this purpose, greyhound racing, a complex domain that involves almost 50 performance variables for eight competing dogs in a race is considered. For every race, each dog's past history is complete and freely available to bettors. This article discusses the experimental procedures as well as the results obtained in the process.
Abstract:
Our society is in a state of transformation toward a "virtual society". However, due to the nature of anonymity and less observability, internet activities have become more diverse and obscure. As a result, unscrupulous individuals or criminals may exploit the internet as a channel for their illegal activities to avoid the apprehension by law enforcement officials. This paper examines the "Gray Web Forums" in Taiwan. We study their characteristics and develop an analysis framework for assisting investigations on forum communities. Based on the statistical data collected from online forums, we found that the relationship between a posting and its responses is highly correlated to the forum nature. In addition, hot threads extracted based on posting activity and our proposed metric can be used to assist analysts in identifying illegal or inappropriate contents. Furthermore, a member's role and his/her activities in a virtual community can be identified by member level analysis. In addition, two schemes based on content analysis were also developed to search for illegal information items in gray forums. The experiment results show that hot threads are correlated to illegal information items, but the retrieval effectiveness can be significantly improved by search schemes based on content analysis. © 2008 Springer-Verlag Berlin Heidelberg.
Abstract:
Nanoscale science and engineering (NSE) have seen rapid growth and expansion in new areas in recent years. This paper provides an international patent analysis using the U.S. Patent and Trademark Office (USPTO) data searched by keywords of the entire text: title, abstract, claims, and specifications. A fraction of these patents fully satisfy the National Nanotechnology Initiative definition of nanotechnology (which requires exploiting specific phenomena and direct manipulation at the nanoscale), while others only make use of NSE tools and methods of investigation. In previous work we proposed an integrated patent analysis and visualization framework of patent content mapping for the NSE field and of knowledge flow pattern identification until 2002. In this paper, the results are updated for 2003, and the new trends are presented. The number of USPTO patents originated from all countries that include nanotechnology-related keywords in 2003 is about 8600, an increase of about 50% over the last 3 years, which is significantly larger than the increase of about 4% for patents in all technology fields (USPTO, 2004). The top five countries are U.S. (5228 patents in 2004), Japan (926), Germany (684), Canada (244) and France (183). Fastest growing are the Republic of Korea (84 patents in 2003) and Netherlands (81). For the first time in 2003, four electronic companies have reached the top five institutions: IBM (198 patents), Micron Technologies (129), Advanced Micro Devices (128), Intel (90) and University of California (89). However, overall, the single technology field "Chemistry: molecular biology and microbiology" and chemical industry remain in the lead. The citation networks show an increase of international interactions, and a relative change of the role of various countries, institutions and technological fields in time.
Abstract:
Concept mapping systems used in education and knowledge management emphasize flexibility of representation to enhance learning and facilitate knowledge capture. Collections of concept maps exhibit terminology variance, informality, and organizational variation. These factors make it difficult to match elements between maps in comparison, retrieval, and merging processes. In this work, we add an element anchoring mechanism to a similarity flooding (SF) algorithm to match nodes and substructures between pairs of simulated maps and student-drawn concept maps. Experimental results show significant improvement over simple string matching with combined recall accuracy of 91% for conceptual nodes and concept → link → concept propositions in student-drawn maps. © 2005 Elsevier B.V. All rights reserved.
