Catálogo de publicaciones - libros

Recently, much attention has been paid to manage amount of personal contents effectively. Private photo is one of the most important types of personal contents. In many of conventional approaches for private photo searching, a user is required to assign annotations manually to the private photos of the user by oneself. However, assigning annotations is tedious for a user. This paper introduces an automatic indexing approach for private photo searching based on e-mail message archive. Many of e-mail messages of a user may describe some experiences of the user. In this approach, the temporal expressions are extracted from an e-mail message, and they are used for estimate the relativity between a term in the e-mail message and a temporal interval in the personal time space of a user. The weight of a term to a private photo is calculated on the basis of the relativity and the granularity of the temporal interval.

This paper proposes an evacuation support method, based on routing panels, and evaluates analytically this method by means of computer simulation. For this simulation, the multi-agent paradigm is adopted as our basic modeling view. On this modeling, we focus on the behaviors of agents promoted instructively by routing panels. As a result, two characteristic suggestions are derived: one is that routing panels set on critical sections, in which were estimated by Monte Carlo method, can support the rapid and safe evacuation; and another is that the route finding way to be applicable by the global disaster information is better than that of the local disaster information.

Automated facial expression classification arises as a difficult, yet crucial, pattern recognition problem in the design of human-computer interaction and multimedia interactive service systems. In the process of building NEU-FACES, a novel system of ours for processing multiple camera images of computer user faces to determine their affective state, we conducted and present here an empirical study in which we specify related design requirements, study statistically the expression recognition performance of humans, and identify quantitative facial features of high expression discrimination and classification power.

Students who enroll in the undergraduate program on informatics at the Hellenic Open University (HOU) demonstrate significant difficulties in advancing beyond the introductory courses. We use decision trees and genetic algorithms to analyze their academic performance throughout an academic year. Based on the accuracy of the generated rules, we analyze the educational impact of specific tutoring practices and reflect on some software engineering issues involved in the development of organization-wide measurement systems.

Intelligent interactive systems have begun to adopt knowledge and software engineering technologies in an attempt to effective development. Feature models have been widely used in knowledge and software engineering for the reuse purpose. However, due to the lack of a formal semantics of feature models, it is rather difficult to perform rigorous consistency reasoning on them. Without guaranteed consistency of feature models, the quality of interactive systems based on them, can not be guaranteed. In this paper, how to formalize feature models with Description Logics is investigated. Following the proposed translation principles, each feature model is formalized into an knowledge base. Hence the consistency reasoning on the feature model turns into the consistency reasoning on the corresponding knowledge base. Especially, the latter reasoning can be automatically performed via the description logic reasoner RACER.

This paper presents a system that integrates news from multiple sources on the Web and delivers in a personalized fashion to the reader. The presented service integrates automatic information extraction from various news sources and presentation of information according to the user’s interests. The system consists of source-specific information extraction programs (wrappers) that extract highlights of news items from the various sources, organize them according to pre-defined news categories and present them to the user through a personal Web-based interface. Dynamic personalization is used based on the user’s reading history, as well as the preferences of other similar users. User models are maintained by statistical analysis and machine learning algorithms. Results of an initial user study have confirmed the value of the service and indicated ways in which it should be improved.

Restricting the form of rules is an important issue of relational association rule mining. The proposing method PIX extracts properties from given examples and to use them to form rules. An property of an instance consists of an addressing part which specifies objects related to the instance and description part which says something among the objects. Extracted properties are used like as an item in market basket database and an APRIORI-like algorithm calculates frequent item sets. The paper describes also an experiment in a sample application.

In the previous studies, authors proposed a new local arc method with new similarity evaluation function to the off-line writer recognition, and obtained high recognition ratios. However, the calculated similarity values are comparatively close each other. Therefore, it is very difficult to improve the accuracy any more. In this study, it is assumed that the amount of the characteristic features appeared to the curvature distribution may be different as the difference of the character size. Then, similarity values of five character sizes were compared, and the optimum character size for the new local arc method was examined. As a result, it turned out that the similarity values are greatly influenced by the character size, and the best character size and arc chord length are obtained for the writer recognition by the new local arc method.

In the previous works, we have developed Web education system using the dynamically changing three layered learning materials and estimation method of understanding level to maintain the eagerness for self learning by changing learning material based on the understanding level. In those studies, the ratio of correct answers was used directly to estimate understanding level. In this study, we developed an understanding level judging system based on the two parameter logistic function obtained from a cumulative frequency distribution and Fuzzy membership functions based on response time and number of explanation referred.

The contents of multimedia data has complex relationships including deeply nested whole-part and the many-to-many relationships. This paper proposes a data model incorporating the concepts of directed graphs, recursive graphs, and hypergraphs in order to represent the contents of multimedia data. In the proposed data model, an instance is represented with a directed recursive hypergraph called an . The logic-based operation called is introduced. It gives us powerful querying capability because regular expressions on paths can be specified in retrieving instance graphs from collection graphs.