Multi-documents sheets are viewed as semantically structured representations of textual documents. The automatic construction of these sheets is based on the automatic annotation of textual documents according to a set of discursive categories called discursive mining viewpoints. The automatic annotation of a text is performed using the Contextual Exploration processing. It is a linguistic and computational method implemented in the EXCOM2 platform that allows the an-notation of segments (which can be a title, a paragraph, a sentence or a clause) according to a given discursive mining viewpoint.