Catálogo de publicaciones - libros

The WWW provides an overwhelming amount of information, which – spatially indexed – can be a valuable additional data source for location-based applications. By manually building a spatial index, only a fraction of the available resources can be covered. This paper introduces a system for the automatic mapping of web pages to geographical locations. Our web robot uses several sets of domain specific keywords, lexical context rules, that are automatically learned, and a hierarchical catalogue of geographical locations that provides exact geographical coordinates for locations. Spatially indexed web pages are used to construct Geographical Web Portals, which can be accessed by different location-based applications. In addition, we present experimental results demonstrating the quantity and the quality of automatically indexed web pages.

Support for temporal text-containment queries is of interest in a number of contexts. In previous papers we have presented two approaches to temporal text-indexing, the V2X and ITTX indexes. In this paper, we first present improvements to the previous techniques. We then perform a study of the space usage of the indexing approaches based on both analytical models and results from indexing temporal text collections. These results show for what kind of document collections the different techniques should be employed. The results also show that regarding space usage, the new ITTX/VIDPI technique proposed in this paper is in most cases superior to V2X, except in the case of patterns of high number of new documents relative to number of updated documents.

We introduce a method of searching the nearest neighbours (-NN) using PM-tree. The PM-tree is a metric access method for similarity search in large multimedia databases. As an extension of M-tree, the structure of PM-tree exploits local dynamic pivots (like M-tree does it) as well as global static pivots (used by LAESA-like methods). While in M-tree a metric region is represented by a hyper-sphere, in PM-tree the ”volume” of metric region is further reduced by a set of hyper-rings. As a consequence, the shape of PM-tree’s metric region bounds the indexed objects more tightly which, in turn, improves the overall search efficiency. Besides the description of PM-tree, we propose an optimal -NN search algorithm. Finally, the efficiency of -NN search is experimentally evaluated on large synthetic as well as real-world datasets.

Adapted from the concept of views in databases, workflow views are derived from workflows as a fundamental support for workflow inter-operability and visibility by external parties in a e-service environment. However, until now there are few works focusing on its realization mechanism, i.e. the communication between views and their source entities. In this paper, we extend the object deputy model to the workflow deputy model supporting the interaction of workflow views in a systematic way. In this workflow deputy model, we formally specify the deputy class and the deputy algebra for workflow classes. According to the process meta-model of XPDL, deputy operations are defined for each kind of workflow component class specifically. Based on this deputy mechanism, workflow views are presented in forms of deputy classes. Lastly, several modeling issues are discussed.

Extracting data from web pages using wrappers is a fundamental problem arising in a large variety of applications of vast practical interests. In this paper, we propose a novel technique to the problem of differentiating roles of data items from Web pages, which is one of the key problems in our automatic extraction approach. The problem is resolved at various levels: semantic blocks, sections and data items, and several approaches are proposed to effectively identify the mapping between data items having the same role. Intensive experiments on real web sites show that the proposed technique can effectively help extracting desired data with high accuracies in most of the cases.

Many data mining techniques which are non-visual methods have been proved their virtues on various customer data. However, there have been hardly applications of visualization methods onto the customer information in spite of their ability of quick and easy knowledge discovery. In this paper, we propose a data visualization method for customer information using a customer map. To develop the customer map, we integrate numerous customer data from various data sources, perform data analyses using data mining techniques and finally visualize the information derived by the former analyses. The customer map makes it possible to mange diverse and complex data sets under the unified goal of value creation through customers. It also affords the ability to make quick observation of current state and the change of customer distribution based on their information without preconception. We applied the customer map to the credit card company, and suggested managerial implications from the customer maps obtained from its data.

A database user session is a sequence of queries issued by a user (or an application) to achieve a certain task. Analysis of task-oriented database user sessions provides useful insight into the query behavior of database users. In this paper, we describe novel algorithms for identifying sessions from database traces and for grouping the sessions different classes. We also present experimental results.

Recovery processing in embedded real-time databases (ERTDBs) is more complex than traditional databases. In this paper, the classifications and consistency constraints of data and transactions in embedded real-time databases are given first. Then time-cognizant recovery principles for different classes of data and transactions are discussed. In terms of these principles, a time-cognizant recovery scheme based on real-time logging is presented, which is suitable for a class of embedded real-time databases applications. Performance evaluations show that the suggested scheme has better performances than traditional recovery techniques in two aspects: the missing deadlines percent of transactions and the time of system denying services after crashes.

In order for a multi-dimensional index structure to be integrated into a commercial database system, efficient concurrency control techniques are necessary. The techniques must support all degrees of isolation offered by the database system. Especially the degree 3 isolation, called no phantom read, protects search ranges from concurrent insertions and the rollbacks of deletions. In this paper, we propose a new phantom protection method for multi-dimensional index structures that uses multi-level grid technique. The proposed mechanism is independent of the types of multi-dimensional index structures, i.e., it can be applied to all types of index structures such as tree-based, file-based and hash-based index structures. Also, it achieves low development cost and high concurrency with low lock overhead. It is shown through various experiments that the proposed method outperforms existing phantom protection methods for multi-dimensional index structures.

Schema matching is a key operation in data engineering. Combining multiple matching strategies is a very promising technique for schema matching. To overcome the limitations of existing combination systems and to achieve better performances, in this paper the CMC system is proposed, which combines multiple matchers based on credibility prediction. We first predict the accuracy of each matcher on the current matching task, and accordingly calculate each matcher’s credibility. These credibilities are then used as weights in aggregating the matching results of different matchers into a combined one. Our experiments on real world schemas validate the merits of our system.