Concept Learning in Text Comprehension

Abstract

This paper presents a mechanism to reverse engineer a cognitive concept association graph (CAG) which is formed by a reader while reading a piece of text. During text comprehension a human reader recognizes some concepts and skips some. The recognized concepts are retained to construct the meaning of the read text while the other concepts are discarded. The concepts which are recognized and discarded vary for every reader because of the differences in the prior knowledge possessed by all readers. We propose a theoretical forward calculation model to predict which concepts are recognized based on the prior knowledge. To demonstrate the truthful existence of this model, we employ a reverse engineered approach to calculate a concept association graph as per the rules defined by the model. An empirical study is conducted of how six readers from an undergraduate class of Computer Networks form a concept association graph given a paragraph of text to read. The model computes a resultant graph which is flexible and can give quantitative insights into the more complex processes involved in human concept learning.