Element Retrieval Using a Passage Retrieval Approach

Element and passage retrieval systems are able to extract and rank parts of documents and return them to the user rather than the whole document. Element retrieval is used to search XML documents and identify relevant XML elements, while passage retrieval is used to identify relevant passages. This paper reports a series of experiments on element retrieval, using a general passage retrieval algorithm. Firstly, an XML document is divided into overlapping or non-overlapping fixed size windows (passages), then the relevant passages which contain query terms are found. Given the position of a passage in the XML document, the smallest element which contains this passage is found. The experiments were conducted with the INEX 2005 ad hoc test collection and evaluation tool. Two passage extraction methods, three weight functions and various window sizes were tested. A comparison with element retrieval systems was also conducted. The experimental results show that a robust passage retrieval algorithm can yield an acceptable level of performance in XML element retrieval.