Merging Textual Knowledge Represented by Element Fuzzy Cognitive Maps

Abstract

Importance degree and difference degree of keywords in different topics have been measured by the associated weights in Element Fuzzy Cognitive Maps (E-FCMs) which can represent textual knowledge effectively. Logic “and” operation is introduced to roughly evaluate the similarities between the mass E-FCMs in order to form the similar sets of textual knowledge. Based on the associated weight measuring and the logic operation, an E-FCMs-based knowledge merging algorithm is proposed to inspect the noisy and the redundancy information hidden in the original E-FCMs belonging to one similar set. A formula obtained through F-measure is employed as an indicator to measure the loss of textual information during the merging process of E-FCMs. The merging algorithm and the indicator provide a concise representation of textual knowledge that can be used in understanding-based automatic text classification and clustering, as well as relevant knowledge aggregation and integration. The proposed algorithm will have very good application prospects in future.