This article presents a proposal for data collection from textual resources in history and the social sciences that we call Computer-Assisted Semantic Text Modelling (CASTEMO). The CASTEMO data model and data collection workflow is based on detailed, yet flexible semantic encoding of the original natural-language syntactic structure and wording: translating texts line by line into structured data while preserving all of their vagaries, complexities, conflicting testimonies and the like. We outline a thorough way of modelling the sources in order to make them accessible to all manner of quantitative and computational analyses.