A local semantic trace is a certain quasi-propositional structure that can still be reconstructed from written content that is incomplete or does not follow a proper grammar. It can also retrace bits of knowledge from text containing only very few words, making the microstructure of these artifacts of knowledge externalization available for automated analysis and comparison--which makes it highly interesting when learners write short texts within or outside of any learning experience. The methodology is designed to track knowledge and understanding in contexts that contain small pieces of speech of this kind, like discussion boards, chats, forums, or any other conceivable discourse, with very small amounts of text. In this paper, the methodology is introduced and cross-validated by two subsequent studies with a total of N = 310 text samples against already existing methods that allow the analysis of medium-length texts with proper grammar. The results show a very promising outlook for three levels of expertise and five completely different domains.