Abstract
The spectacular development of Web-databases and intensive-query applications such as data warehouses (\(\mathcal {DW}\)) contributes to the vulgarization of query-logs. The main particularity of these logs is that they are exploited internally by the owners of these \(\mathcal {DW}\)s. Recently, external resources such as Linked Open Data (\(\mathcal {LOD}\)) bring two main elements: valuable data and stored Sparql query-logs of end users. This situation brings a dual fact: \(\mathcal {DW}\) query-logs are located at the operational \(\mathcal {DW}\), whereas the \(\mathcal {LOD}\) logs are located at a source level. If they are curated and analyzed efficiently, the \(\mathcal {LOD}\) logs may represent an interesting asset for the \(\mathcal {DW}\) design, to build new multidimensional knowledge. In this paper, we propose a \(\mathcal {LOD}\) logs-driven approach for \(\mathcal {DW}\) multidimensional modeling. To show the effectiveness of our approach, we instantiate it using DBpedia query-logs.