<?xml version="1.0" standalone="yes"?>
<Paper uid="W05-0204">
  <Title>Predicting Learning in Tutoring with the Landscape Model of Memory</Title>
  <Section position="3" start_page="0" end_page="21" type="intro">
    <SectionTitle>
2 The Landscape Model
</SectionTitle>
    <Paragraph position="0"> The Landscape Model was designed by van den Broek et al. (1996) to simulate human reading comprehension. In this model, readers process a text sentence-by-sentence. Each sentence contains explicitly mentioned concepts which are added into working memory. In addition, the reader may re-instantiate concepts from earlier reading cycles or from world knowledge in an effort to maintain a coherent representation. Concepts are entered into working memory with initial activation values, which then decay over subsequent reading cycles.</Paragraph>
    <Paragraph position="1"> After concepts are entered, the model calculates connection strengths between them. Two concepts that are active in working memory at the same time will be given a link. The higher the levels of concept activation, the stronger the link will be. Van den Broek et al. (1996) give this formula for calculating link strengths: a0a2a1a3a5a4a7a6a9a8a11a10 a3 a8a13a12 a3 This defines the strength of the connection between concepts x and y as the product of their activations (A) at each cycle i, summed over all reading cycles.</Paragraph>
    <Paragraph position="2"> Two matrices result from these calculations. The first is a matrix of activation strengths, showing all the active concepts and their values for each reading cycle. The second is a square matrix of link values showing the strength of the connection between  each pair of concepts. Van den Broek et al. (1996) demonstrate a method for extracting a list of individual concepts from these matrices in order of their link strengths, starting with the strongest concept.</Paragraph>
    <Paragraph position="3"> They show a correlation between this sequence and the order in which subjects name concepts in a freerecall task.</Paragraph>
    <Paragraph position="4"> In van den Broek's original implementation, this model was run on short stories. In the current work, the model is extended to cover a corpus of transcripts of physics tutoring dialogs. In the next section we describe this corpus.</Paragraph>
  </Section>
class="xml-element"></Paper>