In this paper, we present a workbench for semi-automatic ontology population from textual documents. It provides an environment for mapping the linguistic extractions with the domain ontology thanks to knowledge acquisition rules. Those rules are activated when a pertinent linguistic tag is reached. Those linguistic tags are then mapped to a concept, one of its attributes or even a semantic relation between several concepts. The rules instantiate these concepts, attributes and relations in the knowledge base constrained by the domain ontology. This paper deals with the underlying knowledge capture process and presents the first experiments realized on a real client application from the legal publishing domain.