An Analysis of the Acquired Knowledge

Abstract

Several questions might be posed about the validity of the knowledge acquisition approach presented in this book. It could be argued that LeMICON is only learning statistical correlations and is incapable of learning the structure that is essential to human cognition.¹ This chapter argues that it is possible to extract statistical structural relationships between concepts and organize them into networks of relationships which can support generalization and compilation of knowledge, and predict and explore text-based event sequences. Furthermore, as the results presented so far have suggested, these networks can be used to process and “interpret” text, a form of comprehension. This chapter also includes a general discussion of LeMICON as it compares to other text understanding systems in terms of learning and knowledge acquisition capabilities.