Efficient XML Keyword Search based on DAG-Compression

Abstract: In contrast to XML query languages as e.g. XPath which require knowledge on the query language as well as on the document structure, keyword search is open to anybody. As the size of XML sources grows rapidly, the need for efficient search indices on XML data that support keyword search increases. In this paper, we present an approach of XML keyword search which is based on the DAG of the XML data, where repeated substructures are considered only once, and therefore, have to be searched only once. As our performance evaluation shows, this DAG-based extension of the set intersection search algorithm[1], [2], can lead to search times that are on large documents more than twice as fast as the search times of the XML-based approach. Additionally, we utilize a smaller index, i.e., we consume less main memory to compute the results.