While doing research in digital based context, historians often face the same kind of problem: how to translate their research questions in machine readable format? Find out how to annotate efficiently their material and which techniques/methods to apply are technical steps that tend to become everyday tasks for many Humanities researchers. The main issue is that their questions are too specific to simply apply generic tools and, on the other hand, adapting such tools may require technical abilities that humanists hardly ever have. In such cases, interacting with experts of different backgrounds (e.g. information technologies, developers) could be difficult due to different views, approach to the problem, way of reasoning, etc. In this paper is proposed a methodology, and the research use case on which it is tested, where the close collaboration between historians and engineers allows for a better understanding of the needs of each party, and helps the creation of customizable tools capable of being used in many different contexts and domains. The use-case is to study and compare entities corresponding to the actors of conflicts in a corpus of personal diaries written during World War II. The discourses analyzed are characterized by multiple and very peculiar terminologies, often very ambiguous (like pejorative nicknames). The challenge is to apply generic tools (like named-entity recognition tool – NERD, a POS tagger) and a domain specific dictionary to this corpus, trying not to cross the thin line between generic customization and ad hoc development. Acknowledgement