Ya-Hui Chang

Improving the performance of identifying contributors for XML keyword search

Keyword search is a friendly mechanism for users to identify desired information in XML databases, and LCA is a popular concept for locating the meaningful subtrees corresponding to query keywords. Among all the LCA-based approaches, MaxMatch [9] is the only one which could achieve the property of monotonicity and consistency, by outputting only contributors instead of the whole subtree. Although the MaxMatch algorithm performs efficiently in some cases, there is still room for improvement. In this paper, we first propose to improve its performance by avoiding unnecessary index accesses. We then speed up the process of subset detection, which is a core procedure for determining contributors. The resultant algorithm is called MinMap and MinMap+, respectively. At last, we analytically and empirically demonstrate the efficiency of our methods. According to our experiments, our two algorithms work better than the existing one, and MinMap+ is particularly helpful when the breadth of the tree is large and the number of keywords grows.

A graphical query language for mobile information systems

The advance of the mobile computing environment allows data to be accessed in any place at any time, but currently only simple and ad-hoc queries are supported. People are eager for mobile information systems with more functionality and powerful querying facilities. In this paper, a graphical query language called MoSQL is proposed to be the basis of general mobile information systems. It provides a uniform way for users to access alphanumerical data and to query current or future location information, based on an icon-based interface. The interface is particularly suitable for the mobile environment, since it is easily operated by clicking or dragging the mouse. An example and the underlying theoretical framework will be presented in this paper to demonstrate the functionality of MoSQL.

A Compact and Efficient Labeling Scheme for XML Documents

As XML data nowadays are extensively used in the applications of data exchange and other fields, supporting efficient query processing on XML data, particularly in determining the structural relationships between two elements, is in great demand recently. To avoid the time-consuming tree traversal tasks, many labeling schemes have been proposed to assign each node a unique label, so that the structural relationships between nodes, such as the ancestor-descendant relationship, can be efficiently determined by comparing their labels. However, to the best of our knowledge, none of the existing labeling schemes can support all structural relationships in constant time and also require the least amount of space. In this paper, we propose a labeling scheme based on the concept of the complete tree, which is called the CT (complete-tree) labeling scheme. This labeling scheme is simple and the resultant labels are compact. We formally analyze its properties and perform an empirical evaluation between the CT labeling scheme and other state-of-the-art labeling schemes on different data sets. The experimental results show that the space requirement of our CT labeling scheme is superior to others in most cases. It is also demonstrated that this scheme can efficiently support all structural relationships and may perform even better than other labeling schemes.

Faster algorithms for searching relevant matches in XML databases

Keyword search is a friendly mechanism for the end user to identify interesting nodes in XML databases, and the SLCA (smallest lowest common ancestor)-based keyword search is a popular concept for locating the desirable subtrees corresponding to the given query keywords. However, it does not evaluate the importance of each node under those subtrees. Liu and Chen proposed a new concept contributor to output the relevant matches instead of all the keyword nodes. In this paper, we propose two methods, MinMap and SingleProbe, that improve the efficiency of searching the relevant matches by avoiding unnecessary index accesses. We analytically and empirically demonstrate the efficiency of our approaches. According to our experiments, both approaches work better than the existing one. Moreover, SingleProbe is generally better than MinMap if the minimum frequency and the maximum frequency of the query keywords are close.

Optimizing XML twig queries with full-text predicates

Efficient query processing has been a critical issue for XML repositories. In this paper, we consider the XML query which can be represented as a query tree with twig patterns, and also consists of full-text constraints. Previously, the structure-first approach and the keyword-first approach have been proposed to process such kind of queries. The main focus of this paper is constructing an integrated system to support these two approaches and find the best execution plan. To achieve this goal, we first analyze the components of these two approaches and design a set of operators. We then derive the corresponding cost model and rewriting rules to perform costbased optimization. We also propose several heuristic rules by observing the behaviors of the two approaches. Via an extensive experimental study, we demonstrate that our cost-based system and heuristic system are both effective.

Identifying relevant matches with NOT semantics over XML documents

Keyword search over XML documents has been widely studied in recent years. It allows users to retrieve relevant data from XML documents without learning complicated query languages. SLCA (smallest lowest common ancestor)-based keyword search is a common mechanism to locate the desirable LCAs for the given query keywords, but the conventional SLCA-based keyword search is for AND-only semantics. In this paper, we extend the SLCA keyword search to a more general case, where the keyword query could be an arbitrary combination of AND, OR, and NOT operators. We further define the query result based on the monotonicity and consistency properties, and propose an efficient algorithm to figure out the SLCAs and the relevant matches. Since the keyword query becomes more complex, we also discuss the variations of the monotonicity and consistency properties in our framework. Finally, the experimental results show that the proposed algorithm runs efficiently and gives reasonable query results by measuring the processing time, scalability, precision, and recall.

Locating Valid SLCAs for XML Keyword Search with NOT Semantics

Keyword search provides an easy way for users to pose queries against XML documents, and it is important to support queries with arbitrary combinations of AND, OR, and NOT operators. The previous RELMN algorithm processed such kind of queries by extending the original SLCA definition in a straightforward way, but it did not work correctly in some cases. In this paper, we propose the concept of valid SLCAs as query results. Basically, nodes in an XML document are classified according to their usages, which is further used to define the scope affected by a negative keyword. Only valid nodes, which are not affected by any negative keyword, are qualified to identify valid SLCAs. The experimental results show that the proposed algorithm achieves higher precision and recall, and is more efficient than the previous work.

Efficiently identifying contributors for XML keyword search

Keyword search is a friendly mechanism for users to obtain desired information in XML databases. Among all the research efforts on automatically reasoning the meaningful answers for users, the MaxMatch System [6] proposed the concept of contributor and has attracted a lot of attention. In this paper, we propose the TDPrune system to improve the efficiency of identifying contributors. It is mainly achieved by avoiding processing unnecessary nodes, and also by visiting nodes only once instead of twice. According to the experimental results, our TDPrune system usually outperforms the MaxMatch system, and the difference of execution time is sometimes by an order of magnitude.

Efficient evaluation of XML twig queries

With rapid acceptance of XML technologies, efficient query processing is a critical issue for XML repositories. In this paper, we consider the XML query which can be represented as a query tree with twig patterns. The proposed approach will first quickly retrieve data satisfying fragment paths which consist of only the parent-child relationship, and deal with the ancestor-descendent constraint in the later gluing stage. This approach focuses on where the data need to be “glued” and thus is very efficient. We conduct several experiments to evaluate the performance of the proposed approach. The results show that our system is more efficient than the holistic approach.

Translating XQuery to SQL based on query forests

It is a difficult task to transform an XQuery posed gainst an XML view into an SQL appropriate for the original relational schema to get data. The main reason lies in the difference of the schema modeling power and the query syntax. In this paper, we propose to represent an XQuery as a set of for trees and return trees, or called the query forests as a whole, based on the functionality of the path expressions specified in each clause. These trees show the structural constraint of the input query and serve as the vehicle to combine the mapping SQL fragments into a complete SQL statement. A prototype has been implemented to show the effectiveness of the transformation process.