Search

The purpose and characteristics of e-learning are different from other Web applications. This paper uses an extended Technology Acceptance Model (TAM) to explain the motivation, attitude and acceptance behind participants of e-learning. The factors that determine Web quality include the external variables: system functionality, interface design, pedagogic and contents, as well as community.. Perceived enjoyment was also added to this model as a factor. . The empirical results indicated that the extended TAM explains the acceptability of on-line learning systems with high reliability, validity, and model fitness. All "beliefs" (user perceptions)- perceived usefulness, ease of use, and enjoyment- are good predictors of attitude and acceptance. Pedagogic, and content as well as community are important external factors that predict e-learning acceptance. [ABSTRACT FROM AUTHOR]

One of the main problems of delivering online course materials to learners lies in a deficiency of search capability. We propose a method for querying inside lecture materials exploiting a textbook ontology, which is automatically constructed from a back-of-the-book index and table of contents. The use of a textbook ontology is two-fold: (i) to help users formulate query more efficiently, and (ii) to discriminate query results that are relevant to user information needs from those that merely contain query terms but are not informative. [ABSTRACT FROM AUTHOR]

Predicting future events has great importance in many applications. Generally, rules with predicate events and consequent events are mined out, and then current events are matched with the predicate ones to predict the occurrence of consequent events. Many previous works focus on the rule mining problem; however, little emphasis has been attached to the problem of predicate events matching. As events often arrive in a stream, how to design an efficient and effective event predictor becomes challenging. In this paper, we give a clear definition of this problem and propose our own method. We develop an event filter and incrementally maintain parts of the matching results. By running a series of experiments, we show that our method is efficient and effective in the stream environment. [ABSTRACT FROM AUTHOR]

Some burgeoning web applications, such as web search engines, need to track, store and analyze massive real-time users' access logs with high availability of 24*7. The traditional high availability approaches towards general-purpose transaction applications are always not efficient enough to store these high-rate insertion-only archived streams. This paper presents an integrated approach to store these archived streams in a database cluster and recover it quickly. This approach is based on our simplified replication protocol and high performance data loading and query strategy. The experiments show that our approach can reach efficient data loading and query and get shorter recovery time than the traditional database cluster recovery methods. [ABSTRACT FROM AUTHOR]

The selection and evaluation task of attributes is of great importance for knowledge-based systems. It is also a critical factor affecting systems' performance. By using the genetic operator as the searching approach and correlation-based heuristic strategy as the evaluating mechanism, this paper presents a GA-CFS method to select the optimal subset of attributes from a given case library. Based on the above, the classification performance is evaluated by employing the combination method of C4.5 algorithm with k-fold cross validation. The comparative experimental results indicate that the proposed method is capable of identifying the most related subset for classification and prediction with reducing the representation space of the attributes dramatically whilst hardly decreasing the classification precision. [ABSTRACT FROM AUTHOR]

HiBO is a bookmark management system that incorporates a number of Web mining techniques and offers new ways to search, browse, organize and share Web data. One of the most challenging features that HiBO incorporates is the automated hierarchical structuring of bookmarks that are shared across users. One way to go about organizing shared files is to use one of the existing collaborative filtering techniques, identify the common patterns in the user preferences and organize bookmarked files accordingly. However, collaborative filtering suffers from some intrinsic limitations, the most critical of which is the complexity of the collaborative filleting algorithms that inevitably leads to the latency in updating the user profiles. In this paper, we address the dynamic maintenance of personalized views to shared files from a bookmark management system perspective and we study ways of assisting Web users share their information space with the community. To evaluate the contribution of HiBO, we applied our Web mining techniques to manage a large pool of bookmarked pages that are shared across community members. Results demonstrate that HiBO has a significant potential in assisting users organize and manage their shared data across web-based social networks. [ABSTRACT FROM AUTHOR]

The purpose of this paper is to develop novel Blind Source Separation (BSS) algorithms from linear mixtures of them, which enable to separate dependent source signals. Most of the proposed algorithms for solving BSS problem rely on independence or at least uncorrelation assumption of the source signals. Here, we show that maximization of the nonGaussianity(NG) measure can separate the statistically dependent source signals and the novel NG measure is given by the Hall Euclidean distance. The proposed separation algorithm can result in the famous FastICA algorithm. Simulation results show that the proposed separation algorithm is able to separate the dependent signals and yield ideal performance. [ABSTRACT FROM AUTHOR]

Multicast services can be deployed either on network layer or application layer. Implementations of application-level multicast often provide more sophisticated features, and can provide multicast services where IP multicast services are not available.In this paper, we consider the degree and delay constrained routing problem in overlay multicast for real-time multimedia applications, and an efficient Distributed Tree Algorithm (DTA) is proposed. With DTA, end hosts can make trade-off between minimizing end-to-end delay and reducing local resource consumption by adjusting the heuristic parameters, and then self-organize into a scalable and robust multicast tree dynamically. By adopting distributed and tree-first schemes, a newcomer can adapt to different situations flexibly. The algorithm terminates when the newcomer reaches a leaf node, or joins the tree successfully. Simulation results show that the multicast tree has low node rejection rate by choosing appropriate values of the heuristic parameters. [ABSTRACT FROM AUTHOR]

Web applications have become standard for several areas, however, they tend to be poorly structured and do not have strongly-typed support. In this paper, we present a web application model and a process to extract the model using static and dynamic analysis. We show recurring problems regarding type and structure in web applications and formally describe algorithms to verify those problems. Finally, we show the potentials of our approach via tool support. [ABSTRACT FROM AUTHOR]

Web Communities (WCs) are emerging as business enablers in the electronic marketplace. As the size of the community becomes increasingly large, there is a tendency for some members of the community to use resources provided by the community without necessarily making any contribution. It is, therefore, necessary to determine members' contributions towards sustaining the community. In this paper, we present a formal framework to dynamically measure the interactivity of members, and indeed the interactivity level of the community. This formal foundation is necessary to eliminate ad hoc approaches that characterize existing solutions, and provide a sound foundation for this new research area. We design an efficient interactivity algorithm, and also implement a prototype of the system. [ABSTRACT FROM AUTHOR]

Most agile applications have to deal with dynamic change of processes of automatic business policies, procedures, and logics. As a solution for the dynamic change of processes, rule-based software development is used. With the increase in complexity in modern day business system, the business rules have become harder to express hence require additional especially designed scripting languages. The high cost of modifying or updating those rules is our motivation in this paper. We propose a compilation-based dynamically adjustable rule engine that is used for rich rule expression and performance enhancement. Because of immense complications among and within business rules, we use Java language to create/modify rule instead of scripting languages. It gives us the facility of standardized syntax also. It separates the condition from action during run time which makes rule notification easier and quicker. According to experimental results, the proposed dynamically adjustable rule engine shows promising results when compared with contemporary script-based solutions. [ABSTRACT FROM AUTHOR]

In this paper, a probabilistic graphical modeling approach for Web services is proposed, and the Web services Bayesian network (WSBN) is constructed by mining the historical invocations among them. Further, the semantic guidance to Web services composition is generated based on the Markov blanket and causality reasoning in the WSBN. Preliminary experiments and performance analysis show that our approach is effective and feasible. [ABSTRACT FROM AUTHOR]

Privacy concerns on sensitive data are becoming indispensable in data publishing and knowledge discovering. The k-anonymization provides a way to protect the sensitivity without fabricating the data records. However, the anonymity can be breached by leveraging the associations between quasi-identifiers and sensitive attributes. In this paper, we model the possible privacy breaches as Q-S associations using association and dissociation rules. We enhance the common k-anonymization methods by evaluating the Q-S associations. Moreover, we develop a greedy algorithm for rule hiding in order to remove all the Q-S associations in every anonymity-group. Our method can not only protect data from the privacy breaches but also minimize the data loss. We also make a comparison between our method and one of the common k-anonymization strategies. [ABSTRACT FROM AUTHOR]

The key idea of our k-anonymity is to cluster the personal data based on the density which is measured by the k-Nearest-Neighbor (KNN) distance. We add a constraint that each cluster contains at least k records which is not the same as the traditional clustering methods, and provide an algorithm to come up with such a clustering. We also develop more appropriate metrics to measure the distance and information loss, which is suitable in both numeric and categorical attributes. Experiment results show that our algorithm causes significantly less information loss than previous proposed clustering algorithms. [ABSTRACT FROM AUTHOR]

A Web service architecture is described for bidirectional XML updating. The updating mechanism exploits the power of bidirectional transformation so that users can update remote XML data by editing a view on the local machine that is generated by a transformation of the XML data. This architecture consists of three tiers: data viewer clients, a bidirectional transformation engine, and content servers accessible through the Internet. Due to the use of standard Web service technologies, the data viewer clients and content servers can be easily replaced with ones chosen by the user. Users can use this architecture to implement their own applications that exploit the power of bidirectional transformation without the burden of installing and maintaining a bidirectional language package. [ABSTRACT FROM AUTHOR]

Privacy-preserving data publication for data mining is to protect sensitive information of individuals in published data while the distortion to the data is minimized. Recently, it is shown that (α,k)-anonymity is a feasible technique when we are given some sensitive attribute(s) and quasi-identifier attributes. In previous work, generalization of the given data table has been used for the anonymization. In this paper, we show that we can project the data onto two tables for publishing in such a way that the privacy protection for (α,k)-anonymity can be achieved with less distortion. In the two tables, one table contains the undisturbed non-sensitive values and the other table contains the undisturbed sensitive values. Privacy preservation is guaranteed by the lossy join property of the two tables. We show by experiments that the results are better than previous approaches. [ABSTRACT FROM AUTHOR]

Discovery of frequent XML query patterns in the history log of XML queries can be used to expedite XML query processing, as the answers to these queries can be cached and reused when the future queries "hit" such frequent patterns. In this paper, we propose an efficient bottom-up mining approach to finding frequent query patterns in XML queries. We merge all queries into a summarizing structure named global tree guide (GTG). We refine GTG by pruning infrequent nodes and clustering adjacent nodes in the queries to obtain a Compressed GTG (known as CGTG). We employ a bottom-up traversal scheme based on CGTG to generate frequent query patterns for each node till the root of CGTG. Experiments show that our proposed method is efficient and outperforms the previous mining algorithms of XML queries, such as XQPMinerTID and FastXMiner. [ABSTRACT FROM AUTHOR]

Using index to process structural queries on XML data is a natural way. F&B-Index has been proven to be the smallest index which covers all branching path queries. One disadvantage which prevents the wide usage of F&B-Index is that its construction requires lots of time and very large main memory. However, few works focus on this problem. In this paper, we propose an effective and efficient F&B-Index construction algorithm, SAM, for DAG-structured XML data. By maintaining only a small part of index, SAM can save required space of construction. Avoiding complex computation of the selection of nodes to process, SAM takes less time cost than existing algorithms. Theoretical analysis and experimental results show that SAM is correct, effective and efficient. [ABSTRACT FROM AUTHOR]

Reasoning about the access control model for AXML documents is a non-trivial topic because of its own challenging issues: the hierarchical nature of XML with embedded service call and query transformation. In this paper, we present a methodology to specify an access control model (GUPster) for AXML (Active XML) documents by translating a query, schema, and access control policy in CSP language. Then, we show how to verify access control policies of AXML documents, by illustrating the running example, with the FDR model checker. Finally, the examples demonstrate that our automated static verification is efficient to analyze security problems, not only whether the policies give legitimate users enough permissions to read data, but also whether the policies prevent unauthorized users from reading sensitive data. [ABSTRACT FROM AUTHOR]

Terms used in search queries often have multiple meanings. Consequently, search results corresponding to different meanings may be retrieved, making identifying relevant results inconvenient and time-consuming. In this paper, we propose a new solution to address this issue. Our method groups the search results based on the different meanings of the query. It utilizes the semantic dictionary WordNet to determine the basic meanings or senses of each query term and similar senses are merged to improve grouping quality. Our grouping algorithm employs a combination of categorization and clustering techniques. Our experimental results indicate that our method can achieve high grouping accuracy. [ABSTRACT FROM AUTHOR]

A novel diagnosis method named SeCED-FS is proposed in this paper. The method combines the clusterer ensemble and feature selection technique to improve the diagnosis performance. At first, selective clusterer ensemble with feature selection technique is utilized to perform the classifica- tion of medical images in the two-level architecture. Then, the Regions of Interest in positively identified image are outlined by using an ensemble of Fuzzy C-Means algorithm. Case studies on real data experiments show that, the SeCED-FS holds the improved generalization ability and achieved a satisfac- tory result not only in the accuracy of classification but also correctly labeling the significant regions. [ABSTRACT FROM AUTHOR]

As the deployment of Web services increases in complex business application integration, it is becoming inevitable that several Web services may have the same or similar functionalities each holds different Quality of Services(QoS). In this context, it is becoming a challenge for Web services developers to select most pertinent services for their businesses. In this paper, we suggest a novel method so called Trust Model for selecting provider Web services among several similar providers but with different QoS. Our Trust Model is a function of historically gathered and learned QoS values, references or feedbacks from other services and honesty degree of this service. Client Web services satisfaction degree function was devised for checking the satisfaction degree when using our Trust Model. The proposed model is validated with initial result by comparing two cases: when using our Trust Model for selecting Web services and when not using our Trust Model. [ABSTRACT FROM AUTHOR]

File correlation mining, as a technique to enhance file system performance, can usually be exploited for many purposes such as to improve the effectiveness of cache, to optimize file layout, as well as to enable disk file prefetching. While most research works on file correlations focus on traditional stand-alone file systems, this paper investigates the problem of mining file correlations in a distributed environment. We present a parallel data mining algorithm called PFC-Miner (Parallel File Correlation Miner), which is based on Locality Sensitive Hashing. PFC-Miner can efficiently discover correlations between infrequently-accessed files which are more valuable for web applications. Experimental results show that PFC-Miner can efficiently discover file correlations in distributed file systems without compromising the accuracy, and that the proposed approach has good scalability. [ABSTRACT FROM AUTHOR]

Query transformation is a serious hurdle on semantic peer-to-peer environment. The problem is that the transformed queries might lose some information from the original one, as continuously traveling p2p networks. We mainly consider two factors; i) number of transformations and ii) quality of ontology alignment. In this paper, we propose semantic centrality (SC) measurement meaning the power of semantic bridging on semantic p2p environment. Thereby, we want to build semantically cohesive user subgroups, and find out the best peers for query transformation, i.e., minimizing information loss. We have shown an example for retrieving image resources annotated on p2p environment by using query transformation based on SC. [ABSTRACT FROM AUTHOR]

The heterogeneity nature of XML data creates the need for approximate query answering. In this paper, we present an XML system that cooperates with users to provide user-specific approximate query answering. The key features of the system include: 1) a query language that allows users to specify approximate conditions and relaxation controls; 2) a relaxation index structure, XTAH, that enables the system to provide user-desired relaxations as specified in the queries; and 3) a ranking model that incorporates both content and structure similarities in evaluating the relevancy of approximate answers. We evaluate our system with the INEX 05 test collections. The results reveal the expressiveness of the language, show XTAH's capability in providing user-desired relaxation control and demonstrate the effectiveness of the ranking model. [ABSTRACT FROM AUTHOR]

This paper proposes a method based on Probability Latent Semantic Analysis (PLSA) to analyze web pages that are of interest to the user and the user query co-occurrence relationship, and utilize the latent factors between the two co-occurrence data for building user profile. To make the weight of web pages that user isn't interested decay rapidly, a Fibonacci function is designed as the decay factor for representing the user's interests more exactly. The personalized recommendation is implemented according to the score of web pages. The experimental results showed that our approach was more effective than the other typical approaches to construct user profile. [ABSTRACT FROM AUTHOR]

Collaborative Filtering recommender system, one of the most representative systems for personalized recommendations in E-commerce, is a system assisting users in easily finding useful information. However, traditional collaborative filtering systems are typically unable to make good quality recommendations in the situation where users have presented few opinions; this is known as the cold start problem. In addition, the existing systems suffer some weaknesses with regard to quality evaluation: the sparsity of the data and scalability problem. To address these issues, we present a novel approach to provide enhanced recommendation quality supporting incremental updating of a model through the use of explicit user feedback. A model-based approach is employed to overcome the sparsity and scalability problems. The proposed approach first identifies errors of prior predictions and subsequently constructs a model, namely the user-item error matrix, for recommendations. An experimental evaluation on MovieLens datasets shows that the proposed method offers significant advantages both in terms of improving the recommendation quality and in dealing with cold start users. [ABSTRACT FROM AUTHOR]

The increasing copies of digital documents make detecting duplicates an important problem. Among the techniques proposed so far, Winnowing fingerprinting algorithm [5] is one of the most efficient. However, the previous density analysis leave the performance of Winnowing unwarranted in real systems, because the assumption of uniformly distributed k-grams is far from true in practice. In this paper, an improved density analysis method is introduced. Compared with the previous, our method needs only identically distributed k-grams to get the prediction. This means our theoretical result can be safely used on highly non-uniformly distributed data which are common in real systems. Extensive experiments are performed on both artificial data and real data. The experiment results agree with the theoretical predictions well. [ABSTRACT FROM AUTHOR]

In recent years, the research of record extraction from large document data is becoming popular. However there still exist some problems in record extraction. 1) when large document data is used for the target of information extraction, the process usually becomes very expensive. 2) it is also likely that extracted records may not pertain to the user's interest on the aspect of the topic. To address these problems, in this paper we propose a method to efficiently extract those records whose topics agree with the user's interest. To improve the efficiency of the information extraction system, our method identifies documents from which useful records are probably extracted. We make use of user feedback on extraction results to find topic-related documents and records. Our experiments show that our system achieves high extraction accuracy across different extraction targets. [ABSTRACT FROM AUTHOR]

If the user wants to know trustworthiness of a proposition, such as whether gthe Japanese Prime Minister is Junichiro Koizumih is true or false, conventional search engines are not appropriate. We therefore propose a system that helps the user to determine trustworthiness of a statement that he or she is unconfident about. In our research, we estimate trustworthiness of a proposition by aggregating knowledge from theWeb and analyzing creation time of web pages. We propose a method to estimate popularity from temporal viewpoint by analyzing how many pages discussed the proposition in a certain period of time and how continuously it appeared on the Web. [ABSTRACT FROM AUTHOR]

Continuous K nearest neighbor (CkNN) queries under mobile environment are defined as the k nearest neighbors search for a query object at a serial query time, and all the query object and the target objects are moving. In this paper, we propose method to solve this problem, focusing on finding out the relations between the continuous queries, and proposing decision method for relative moving trend among moving objects. Our experiments show that our approach outperforms the original straightforward method, specially when the query interval is small. [ABSTRACT FROM AUTHOR]

We study the problem of segmenting time series stream. Existing segmenting methods for time series mainly focus on the static data, and may be infeasible under the circumstance of time series stream. We propose an approximate method of APCAS(Adaptive Piecewise Constant Approximate Segmentation) to adaptively segment time series stream, which works in linear time. Extensive experiments, both on synthetic and real datasets, show that our approach is efficient and effective. [ABSTRACT FROM AUTHOR]

Group Nearest Neighbor query is a relatively prevalent application in spatial databases and overlay network. Unlike the traditional KNN queries, GNN queries maintain several query points and allow aggregate operations among them. Our paper proposes a novel approach for dealing with difference operation of GNN queries on multiple query points. Difference nearest neighbor (DNN) plays an important role on statistical analysis and engineer computation. Seldom existing approaches consider DNN queries. In our paper, we use the properties of hyperbola to efficiently solve DNN queries. A hyperbola divides the query space into several subspaces. Such properties can help us to prune the search spaces. However, the computation cost using hyperbola is not desirable since it is difficult to estimate spaces using curves. Therefore, we adopt asymptotes of hyperbola to simplify the hyperbola-based pruning strategy to reduce the computation cost and the search space. Our experimental results show that the proposed approaches can efficiently solve DNN queries. [ABSTRACT FROM AUTHOR]

With expanding of query requirements, event-relative semantic has appeared for data stream applications. This paper builds a simple but effective event-driven data stream model EQM based on event semantics and features over data stream. Furthermore, some improved approaches over event detection and queries as well as relative efficiency evaluation are discussed. Experiments show that our approaches gain better performance than available data stream model and processing approaches as far as event-relative problem is concerned. [ABSTRACT FROM AUTHOR]

The increasing demand on streaming data processing has motivated the study of providing Quality of Service (QoS) for data stream processing. Especially for mission-critical applications, deterministic QoS requirements are always desired in order for the results to be useful. However, the best-effort QoS providing of most existing data stream systems would bring about considerable uncertainty to the query results. This paper attempts to provide a new insight into the problem of providing QoS guarantee for continuous queries over streams. Based on the proposed QoS model of stream processing, a QoS guaranteeing scheduling algorithm is proposed. Experimental results are presented to characterize the efficiency and effectiveness of our approaches. [ABSTRACT FROM AUTHOR]

More and more applications involve processing continuous data streams, and the data stream management system (DSMS) is designed to deal with such data streams. Due to features of large volume and stochastic arrival, DSMS must process data stream efficiently in order to avoid system memory exhaustion and reduce the data access latency, to satisfy requirements of the application requirement. One of the key factors, which significantly impact the system performance significantly, is the scheduling strategy adopted by the DSMS. Chain scheduling is an operator-based scheduling strategy for DSMS, which has near-optimal in terms of run-time memory usage. FIFO strategy achieves optimal performance in terms of data access latency. Inspired by the two important scheduling strategies, Chain and FIFO, we propose two novel adaptive strategies for DSMS, ASCF and CSS, which efficiently deal with the varying input load in terms of both memory usage and data access latency. To give a fair comparison performance with other competing strategies, we design thorough simulation experiment and run different strategies under the same system environment. The outcomes of simulation experiment demonstrate the potential benefits and advantages of ASCF and CSC. [ABSTRACT FROM AUTHOR]