Catálogo de publicaciones - libros

This paper introduces our method for causal knowledge retrieval from the Internet resources, its results and evaluation of using it in utterance creation process. Our system automatically retrieves commonsensical knowledge from the Web resources by using simple web-mining and information extraction techniques. For retrieving causal knowledge the system uses three of specific several Japanese “if” forms. From the results we can conclude that Japanese web pages indexed by a common search engine spiders are enough to discover common causal relationships and this knowledge can be used for making Human-Computer Interfaces sound more natural and interesting than while using classic methods.

This research investigates continuous forecasting queries with alterable forecasting-step over data streams. A novel Adaptive Precision forecasting method to forecasting single attribute value of item in a single steam (stream-value), called AForecast, is proposed. The concepts of dual slide windows and forecasting-steps conduction operator are introduced. AForecast determines multiple forecasting-step based on the change ratio of stream-value and forecasts random-variant stream-value using relative precise prediction of deterministic components of data streams. Based on the theory of interpolating wavelet and optimal linear kalman filtering, this method can approximately generate optimal forecasting precision. Experiment results on actual power load data prove that this method can provide online accurate prediction on stream-value.

Closed itemset mining is a difficult problem especially when we consider the task in the context of a data stream. Compared to mining from a static transaction data set, the streaming case has far more information to track and far greater complexity to manage. In this paper, we propose a complete solution based on CLOSET+ algorithm to closed itemset mining in data streams. In data streams, bounded memory and one-pass constraint are expected. In our solution, these constraints are both taken into account.

Schema matching, the problem of finding semantic correspondences between elements of source and warehouse schemas, plays a key role in data warehousing. Currently, the mappings are largely determined manually by domain experts, thus a time-consuming process. In this paper, based on a multistrategy schema matching framework, we develop a linguistic matching algorithm using semantic distances between words to compute their semantic similarity, and propose a structural matching algorithm based on semantic similarity propagation. After describe our approach, we present experimental results on several real-world domains, and show that the algorithm discovers semantic mappings with a high degree of accuracy.

The amount of computing time for K Nearest Neighbor Search is linear to the size of the dataset if the dataset is not indexed. This is not endurable for on-line applications with time constraints when the dataset is large. However, if there are categorical attributes in the dataset, an index cannot be built on the dataset. One possible solution to index such datasets is to convert categorical attributes into numeric attributes. Categories are ordered and then are mapped to numeric values. In this paper, we propose a new heuristic ordering algorithm to compare with two previously proposed algorithms that borrow the idea from minimal spanning trees. The new algorithm divisively builds a binary tree by recursively partitioning the categories. Then, we in-order traverse the tree and get an ordering of the categories. After mapping and indexing, we can efficiently retrieve a small portion of the dataset and perform K nearest neighbor search on the portion at the cost of a little bit of accuracy. Experiments show the divisive ordering algorithm performs better than the other two algorithms.

Buyers always want to obtain goods at the lowest price. To achieve this, a buyer agent can have multiple concurrent negotiations with all the sellers. It is obvious that when the buyer obtains a new low offer from one of the sellers during negotiation, the buyer should have more bargaining power in negotiating with other sellers, as other sellers should offer a even lower price in order to make a deal. In this way, the concurrent negotiations mutually influence one another. In this paper, we present an algorithm to enable mutual influence among multiple concurrent negotiations. We show that the outcome of this algorithm is in Nash equilibrium when we model and analyze the negotiation as a game.

Personal identification technology is a critical field of information security that has raised various equipments and communications in the modern society. Most studies have concentrated on fingerprint input sensors, recognition, and performance evaluation. However, the vulnerability evaluation for integrity verification of fingerprint recognition systems has partly been studied so far. In this paper, we design and implement two prototype evaluation tools for the vulnerability evaluation of a matching algorithm and integrity evaluation in the fingerprint recognition. We show the results of integrity violation through the process of normal transfer and variant transfer for extracted minutiae data.

Model updating, as a new concept to be employed as a standard and universal method for system modification, has been started in [4] and further developed in this paper. This paper introduces algorithms for Computation Tree Logic (CTL) model update. The algorithms correct errors in a CTL Kripke model to make this model satisfy its various required properties. These update algorithms are designed through pseudo-code, which details the logic of the algorithms. The microwave oven example in [3] has been correctly treated by the pseudo-code, which demonstrates the feasibility of these algorithms. The algorithms and their pseudo-code are the foundations for later implementation of a CTL model updater and integration of a CTL model checker and this updater.

The new market forces have affected the operation of distribution centers (DCs) extremely. The DCs demand an increased productivity and a low cost in their daily operation. This paper focuses on the order-picking operation, which is the core operation of DCs, to improve the material-handling process and enhance the competitiveness. The study applied an integrated framework that combined two algorithms, branch-and-bound and tabu search, to the order-picking problem and developed a neighborhood selection criterion based on constraint propagation and the prune-and-search algorithm to improve computational efficiency. The proposed method forbids bad node when searching a solution by branch-and-bound algorithm or tabu search. Besides, in order to make the algorithm more robust and reasonable, the fuzzy modeling was employed. The empirical result shows the approach decreases the computational effort effectively on the both algorithms. The study may be of interest for the researchers in the optimization of DCs’ operations and the DCs’ managers.

In order to solve multidimensional scaling (MDS) efficiently, we proposed an algorithm, which apply stochastic gradient algorithm to minimizing well-known MDS criteria [1]. In this paper, the efficient MDS algorithm is applied to the text mining and compared with the SOM [2]. The results verified the validity of our algorithm in the analysis of a massive document collection. Our algorithm could find out some interesting structures from about 100000 articles in Usenet (NetNews).