RICHARD GOLDEN HOME PAGE

Text Comprehension and Memory Research


When a reader reads a text, an "understanding" of the text is created in the reader's mind. It is convenient to introduce the technical term "situation model" in order to refer to this understanding of a text which is created in the mind of a reader. The process of constructing a situation model is called the "comprehension process", while operations such as "text recall", "text summarization", and "text question-answering" generate scientifically observable events as a by-product of the interactions of the situation model and other knowledge structures in the reader's mind.
Dr. Golden is currently developing a statistical model which allows the user to incorporate prior knowledge about the semantic and syntactic relationships among the elements in a text. This statistical model is called Knowledge Digraph Contribution (KDC) Analysis. The parameters of the model can be estimated from human recall, summarization, and question-answering data. Each parameter may be interpreted as the relative strength of a different knowledge schema. The parameters of the KDC model are uniquely determined and the large sample probability distribution of the estimates of the parameters can be derived for large sample sizes. These mathematical results are relevant for deriving customized statistical tests for testing hypotheses about the relevance of specific knowledge schema weighting parameters. Statistical tests for deciding which of several text knowledge schemata "best-fits" a given set of recall data have also been developed. More recently, methods of sampling from the KDC probability model have been derived which allow one to generate synthetic recall protocols and then compare these synthesized recall protocols with actual human recall protocols for the purpose of evaluating the validity of the user-specified "knowledge analyses" of the text.

Dr. Golden and his graduate students are also working on the AUTOCODER project which is a software tool that facilitates the coding of recall, summarization, talk-aloud, and question-answering protocol data.

RICHARD GOLDEN HOME PAGE