Collation is one of the first philological activities for which the use of computers has been considered. Already during the 60’s, algorithms for semi-automatic collation came to light. This article pursues their development, focusing on a specific type of materials: medieval vernacular texts. The linguistic resources available nowadays for these materials allow to design a collation pipeline, following the architecture proposed by the Gothenburg model: the text of each witness is annotated with linguistic information; the alignment is made on the lemma, in order to neutralize the orthographic variation; eventually, the linguistic annotation is used to identify categories of variants. While most of the steps in this pipeline have been conceived earlier in the history of semi-automatic collation, they were partially carried out manually; the possibility of automatize them might influence the adoption of computers for the collation of medieval vernacular texts.