Identifying new knowledge in texts through corpus analysis

Taking knowledge as comprising concepts and conceptual associations, this paper attempts to identify knowledge likely to be new to readers in an informative text through a corpus analysis based on lexical priming theory. Potential new concepts are identified through a keyness comparison between the text and the British National Corpus (BNC), taken as a rough representation of readers’ likely existing knowledge. Potential new conceptual associations are identified through a comparative z-score analysis of wide-span co-occurrences in the text and the BNC. This approach appears to have potential and has applications in text mining.