Relaxing XML Preference Queries for Cooperative Retrieval SungRan Cho and Wolf-Tilo Balke, L3S Research Center, Leibniz University of Hannover, 30167 Hannover, Germany {scho, balke}@L3S.de Abstract. Today XML is an essential technology for knowledge management within enterprises and dissemination of data over the Web. Therefore the effi- cient evaluation of XML queries has been thoroughly researched. But given the ever growing amount of information available in different sources, also query- ing becomes more complex. In contrast to simple exact match retrieval, approx- imate matches become far more appropriate over collections of complex XML documents. Only recently approximate XML query processing has been pro- posed where structure and value are subject to necessary relaxations. All the possible query relaxations determined by the user's preferences are generated in a way that predicates are progressively relaxed until a suitable set of best possi- ble results is retrieved. In this paper we present a novel framework for develop- ing preference relaxations to the query permitting additional flexibility in order to fulfil a user’s wishes. We also design IPX, an interface for XML preference query processing, that enables users to express and formulate complex user pre- ferences, and provides a first solution for the aspects of XML preference query processing that allow preference querying and returning ranked answers. Keywords: XML query processing, preference-based retrieval, personalization. 1 Introduction XML is widely used as a base technology for knowledge management within enterprises and dissemination of data on the Web (like e.g., product catalogues), because it allows to organize and handle semistructured data. A collection of XML documents is viewed as a forest of node labeled trees. The data generally can be queried on both structure and content using advanced retrieval languages such as XPath or XQuery. User queries will usually be structured to express the user's information needs and users often have quite specific preferences about the structure, especially when the document structure shows a certain semantics often described by DTDs or XML schema. However, due to the large number and complexity (or heterogeneity) of XML documents, the retrieval process should be cooperative between system and user. Here approximate matches that allow to rank answers according to their relevance to the query [3, 6, 19, 20] are more appropriate than exact match queries. Recently several proposals have therfore studied ranking methods that account for structure to score answers to XML queries.