Creation of knowledge-added concept maps: time augmention via pairwise temporal analysis

Purpose: Although acknowledged as a principal dimension in the context of text mining, time has yet to be formally incorporated into the process of visually representing the relationships between keywords in a knowledge domain. This paper aims to develop and validate the feasibility of adding temporal knowledge to a concept map via pair-wise temporal analysis (PTA). Design/methodology/approach: The paper presents a temporal trend detection algorithm – vector space model – designed to use objective quantitative pair-wise temporal operators to automatically detect co-occurring hot concepts. This PTA approach is demonstrated and validated without loss of generality for a spectrum of information technologies. Findings: The rigorous validation study shows that the resulting temporal assessments are highly correlated with subjective assessments of experts (n = 136), exhibiting substantial reliability-of-agreement measures and average predictive validity above 85 per cent. Practical implications: Using massive amounts of textual documents available on the Web to first generate a concept map and then add temporal knowledge, the contribution of this work is emphasized and magnified against the current growing attention to big data analytics. Originality/value: This paper proposes a novel knowledge discovery method to improve a text-based concept map (i.e. semantic graph) via detection and representation of temporal relationships. The originality and value of the proposed method is highlighted in comparison to other knowledge discovery methods.