With the growth of textual data, techniques are necessary for their selection and organization. The organization of textual documents belonging to specialized languages relies mainly on information reflecting the subject domain, the genre of the document and its communication level. This paper aims to distinguish different levels of specialized documents: scientific documents and popular science documents. Identifying the communication level of a document requires a modality analysis that characterizes the relationship between the author and the content. This paper investigates the relevancy of two modality theories defined for general language in order to characterize texts in the medical field. Furthermore, it compares these theories with respect to their multilingual adaptation working on French and Japanese. It shows that the automation of both theories using specific features is possible, but because of paucity and ambiguity problems, only one is fully operational and scales efficiently.