Named Entity Based Ranking with Term Proximity for XML Retrieval

In this article, an indexing scheme that includes the named entity category for each indexed term is proposed. Based on this, two methods are proposed, one to infer the semantics of an XML element based on its data content, called the confidence value of the element, and the second method computes the proximity scores of the query terms. The confidence value of an element is obtained based on the probability of a named entity category in the data content of the underlying XML element. The proximity score of the query terms measures the proximity and ordering of the query term within an XML element. The article then shows how a ranking function uses the confidence value of an XML element and proximity score to mitigate the impact of higher frequency terms and compute the relevance between a keyword query and an XML fragment. Finally, a keyword search system is introduced and experiments show that the proposed system outperforms existing approaches in terms of search quality and achieve a higher efficiency.