| XML (eXtensible Markup Language) is a set of semantic markup standard defined by W3C. With the rapid development of network applications, the data on web increase exponentially. XML is becoming the standard format on the Internet for data exchange and description information, and widely applied to the digital libraries, data integration, Web Service and so on. XML type data is becoming primary form of data on the network. So information retrieval in the XML data is becoming an important research direction.Compared to XML structured query, XML keyword query becomes an important branch of the XML data retrieval. XML keyword search methods are based on the formation of the LCA (Lowest Common the Ancestor). In order to improve query quality and efficiency, researchers also proposed a number of query semantics, including the SLCA (Smallest LCA), MLCA (Meaningful LCA), VLCA (Valuable LCA) and so on. There may be a lot of problems in practical applications for these query semantics and implement algorithms, such as return meaningless results, loss meaningful information and so on.In this paper, we consider an XML document as a collection of entities, attribute, value in the real world, which is similar to the ER model in relational database. Different elements in the XML document represent different types of information. By considering entity as the basic semantic unit, we propose a new semantics: Lowest Common Entity Ancestor (LCEA) to answer keyword search over XML documents. A LECA node in an XML document represents complete information unit of the real world. Based on LCEA, we propose concept of Smallest Lowest Common Entity Ancestor (SLCEA) semantic. It solves the problem with result not completely and meaningless, meanwhile provides easy-to-use query ways for users.On the basis of the entity concept, we improve search method according to the relationship between the elements. It not only supports unambiguous query, and also supprots for ambiguous query and advanced search. Fisrt we partition keyword query according to different types of entities, and then filter the inverted list of entities partition, lastly adopt ILE algorithm to calculate the smallest lowest common entity ancestor (SLCEA).When output results, we infer different outputting information according to the different case of the LCEA. For ambiguous query processing, our approach can identify and rank different search intention, and return the results of each search intention. This method is more convenient for users to find useful information. The last experience verifies the performance of our approach in search quality and implementation efficiency. |
