Natural Language Processing and XML Retrieval
Alan Woodley, Marcus Hassler, Universitdat Klagenfurt, Xavier Tannier, Shlomo Geva · QUT ePrints (Queensland University of Technology) · 2006
Traditional information retrieval systems respond to users’ queries with a ranked list of relevant documents. XML documents differ from flat text documents by explicitly separating content and structure. XML-IR systems aspire to exploit the additional resources of XML documents by providing users with more specific information than at the document level. However, in order to interact with XML-IR systems users require an interface that is powerful enough to capture their content and structural information need while also been user friendly. The standard interface for document-level IR systems are keywords, however, these only capture users’ content needs. Historically, formal languages, such as the XPath-like NEXI, have been used for XML-IR, however, they have proven too difficult to use by experts in XML-IR, let alone users less familiar with the domain. Natural language interfaces provide a novel approach that is both powerful enough to full capture users’ content and structural needs and user-friendly enough that it can be used intuitively. Furthermore, the success of applying natural language in information seeking fields that focus on a finer granularity of information than documents (such as question and answering) showcases the exciting potential for end-users if the fields of NLP and XMl-IR collaborate.