Faïza Abbaci, Pascal Francq

In this paper, we consider the problem of ranking XML components according to their relevance to a given query. As
opposed to ranking documents, ranking XML components introduces new criteria to take into account in determining the degree
of relevance of a component, such as the granularity of a component or the proximity of query keywords within a component. Most
proposed approaches combine the relevance criteria values in a single function to compute the relevance score for each component
retrieved. The major disadvantage of using a single global function is that good values of some criteria compensate bad ones. In
order to overcome this problem, we propose to use a multi-criteria decision support method called PROMETHEE to rank the XML
components. PROMETHEE is based on a principle of pairwise comparisons between the components with regard to each criterion.
We present the experiments conducted to evaluate the adequacy of the criteria selected and the performance of our approach.