| With the rapid expansion of the World Wide Web, more and more databases that are available online and accessible only via Web form based interfaces are emerged over the Web, these databases are referred to as Web databases. In recent years, with the universal use of the Internet and fast grows of the size of Web databases, accessing the Web database has become an important way for people to obtain the information.The existing Web database query processing models have usually assumed that users know what they want and they supported only a strict query matching model. But with the increasing of the scope and complexity of the Web database, it is unrealistic to make lay users to know the database structure and contents, and so that too little even empty answers may be returned from a Web database in response to a user query even if it is explicit. In such a context, to relax the original query for presenting more relevant answers is desired to the lay users especially the users demanding "instant gratification". After the query relaxation, however, users may be confronted with many answers problem, and it will be desirable to have the option of ordering or categorizing the matches automatically in order to deal with information overload. Moreover, many users usually have vague of imprecise ideas when searching the Web database and the queries they submitted may only be fuzzy descriptions of their query intentions, thus it is desirable to support the expression of fuzzy queries and to search the Web database using fuzzy predicates directly. From the above, it is not difficult to see that the users’ expectation of solving the problems above reflects the need of flexible querying service the Web database systems should provide, while the existing Web database processing models can satisfy such needs neither in the aspect of the query expression nor in the aspect of query processing.In this dissertation, the problems of empty answers, many answers and fuzzy query, which occur in searching the Web databases and standing in need of solutions, are investigated. Also, from the perspective of satisfying the users’needs and preferences, an efficient flexible query solution and corresponding technologies for the Web database, in accordance with the order of query relaxation, relaxed query results and categorization and fuzzy query, are proposed. The main contributions of this dissertation are summarized as follows:(i) To deal with the problem of empty answers of the Web database, an adaptive query relaxation approach, which is based on semantic similarity, is proposed. Firstly, according to the query conditions and data distribution the importance of each specified attribute for the user is speculated, and then an attribute weight measuring method is proposed. Next, based on the properties of attribute values, the semantic similarity measuring methods of categorical attribute values (resp. numerical attribute values) are proposed. According to the relaxation threshold, attributes weights and semantic similarities of attribute values, an adaptive query relaxation rewriting algorithm is proposed, and a method that takes advantages of the satisfaction degree of tuples and initial query to rank the relevant answers is presented as well. Results of experiments demonstrate that the performance and results of attribute weight and attribute values similarity measuring methods proposed are stable and reasonable respectively, the query relaxation method proposed has higher recall and can capture the user’s needs and preferences more effectively as well.(ii) To deal with the problem of many answers returned from a Web database in response to a relaxed query, a contextual preferences-based query results ranking approach is proposed. Firstly, a model of contextual preference with interest degree which can embody both the preference relation and preference degree is presented by combining the representations of qualitative and quantitative preferences. And then, the obtaining and processing methods of contextual preferences with interest degrees are presented, respectively. Based on the contextual preferences with interest degrees, an approach for ranking many answers of a relaxed query is proposed. Results of experiments demonstrate that the preference model proposed has a strong preference expressive ability, and the ranking method proposed also has higher ranking quality and execution efficiency.(iii) A categorization approach, which is complementary to the ranking approach and used for categorizing the many answers of Web database, is proposed. Firstly, based on the vector space model, a method for measuring the similarities of different queries is proposed, and then a method for grouping the similar queries in query history and a method for clustering database tuples based on the queries groups are proposed, respectively. Next, based on the tuples clusters and modified C4.5 decision tree categorization algorithm, a categorization tree construction method for query results is presented. Results of experiments demonstrate that the results of queries similarity measuring are reasonable, and the categorization method proposed also has better categorization effectiveness and lower searching cost,(iv) To deal with the fuzzy query problem of the Web database, a knowledge-based fuzzy query translation and results ranking approach is proposed. Firstly, based on the fuzzy sets theory, a method which synthesizes the membership functions, domain knowledge, weighting functions and theα-cut operation for realizing the fuzzy query translation, is proposed. This process procedure fully considers the importance of each fuzzy basic query condition to the user. And then, according to the satisfaction degree of result tuples to the fuzzy query and user preferences, two ranking methods for the Web database fuzzy query results are presented, respectively. Results of experiments demonstrate that the fuzzy query method proposed achieves both higher recall and precision, and has higher execution efficiency as well. |
PDF Full Text Request