Local Semantic Trace: A Method to Analyze Very Small and Unstructured Texts for Propositional Knowledge

• In Event: 29.088 - Measurement and Assessment With Technologies for Learning

Abstract

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 that contains only very few words, making the microstructure of such artifacts of knowledge externalization available to automated analysis. The methodology is built to track knowledge and understanding in contexts that contain such small pieces of speech, like discussion boards, chats, forums. In this paper, the methodology is introduced and cross-validated by two studies with a total of N=310 text samples against already existing methods. The results show a very promising outlook for three levels of expertise and five completely different domains.

Author

• Pablo Nicolai Pirnay-Dummer, University of Halle, Germany