The oversupply of data, information and knowledge, even after preliminary keywords and topics search, is a well-known problem in R&D activities. One of the approaches aimed at limiting the negative impact of the surplus of information is its automated intelligent preprocessing and reuse. The paper describes a method for identification of the concepts which is based on combination of subject-driven document clustering, shape analysis, trends understanding and relevant context retrieval via semantic analysis. The goal is to extract potentially interesting knowledge from a set of documents based on analysis of graphical information and next to explain the mechanism of the studied process. The proposed method is implemented in the software suite which contains source searching tool, plot comparator and semantic analyzer. The method has been applied to identify the calculation process, using channel geometry characteristics, of the longitudinal dispersion coefficients for one branch of the Somes river in Romania. (C) 2008 Elsevier Ltd. All rights reserved.