The work on the named entity recognition (NER) in databases of historical texts has been placed among the most promising new ways to implement best recovery and managements tools for exploring mass data. In this paper, we describe the application processing NER through a modelling with CRF on an annotated database of Burgundy collection of charters from the tenth to thirteenth centuries. The aim is to generate a model for automatic recognition of named entities in historical sources. We discuss the nature of historical documents in the corpus and extraction of rules, and we expose adaptation to the processing algorithm and the most common problems encountered in Medio Latin texts using diplomatic formularies, which is an atypical case within the NER studies.