The purpose of this study is to develop a better understanding of technologies that use natural language as the basis for concept map construction. In particular, this study focuses on the semantic relation (SR) approach to drawing rich and authentic concept maps that reflect students' internal representations of a problem situation. The following discussions are included: (a) elaborate classifications of concept map approaches that use natural language responses (e.g., student essay); (b) the SR process of eliciting concept maps, established using studies on domain ontology; and (c) a more effective way to identify key concepts and relations from a concept map generated by the SR approach. By comparing the SR approach to other promising concept map technologies that constrain the analytical process in various ways, this study suggests that the SR approach is likely to draw richer and more authentic concept maps. In addition, this study suggests that a certain combination of graph-related metrics be used to filter key concepts from a SR concept map drawn from a written text of 350-400 words. The methods suggested in the study could be used to design an automated assessment technology for complex problem solving and to develop adaptive learning systems.