<?xml version="1.0" standalone="yes"?>
<Paper uid="W99-0409">
  <Title>Exploiting the Student Model to Emphasize Language Teaching Pedagogy in Natural Language Processing</Title>
  <Section position="1" start_page="0" end_page="55" type="abstr">
    <SectionTitle>
Abstract
</SectionTitle>
    <Paragraph position="0"> One of the typical problems of Natural Language Processing (NLP) is the explosive property of the parser and this is aggravated in an Intelligent Language Tutoring System (ILTS) because the grammar is unconstrained and admits even more analyses. NLP applications frequently incorporate techniques for selecting a preferred parse. Computational criteria, however, are insufficient for a pedagogic system because the parse chosen will possibly result in misleading feedback for the learner. Preferably, the analysis emphasizes language teaching pedagogy by selecting the sentence interpretation a student most likely intended. In the system described in this paper, several modules are responsible for selecting the appropriate analysis and these are informed by the Student Model. Aspects in the Student Model play an important pedagogic role in determining the desired sentence interpretation, handling multiple errors, and deciding on the level of interaction with the student.</Paragraph>
    <Paragraph position="1"> Introduction One of the fundamental problems of any Natural Language Processing (NLP) system is the often overwhelming number of interpretations a phrase or sentence can be assigned. For example, van Noord (1997) states that the Alvey Tools Grammar with 780 rules averages about 100 readings per sentence on sentences ranging in length between 13 and 30 words. The problem is not always improved with deeper analysis, for though a semantic analysis may rule some of the possible syntactic structures, it will introduce lexical and scope ambiguity.</Paragraph>
    <Paragraph position="2"> The problem of resolving multiple interpretations is compounded in an Intelligent Language Tutoring System (ILTS) because the grammar must not only admit grammatical structures, but must also be able to navigate over ungrammatical structures and record the errors that the student has made. As a consequence, a grammar for an ILTS will not only assign structures to a grammatical sentence, but may also find analyses which interpret the sentence as ungrammatical, a set of analyses that a traditionally constrained grammar would not find.</Paragraph>
    <Paragraph position="3"> The usual method of limiting the number of parses that an ILTS grammar assigns is to examine the effects of relaxing those constraints that represent likely sources of error by students and introduce new constraints into the grammar rules to block unlikely parses (Schneider &amp; McCoy 1998). Such techniques, however, overlook individual learner differences as a key factor in language teaching pedagogy.</Paragraph>
    <Paragraph position="4"> The system introduced in this paper differs from the traditional approach by permitting the grammar to freely generate as many parses as it can and using separate pedagogic principles to select the appropriate interpretation and response. The system tightly integrates the Student Model into the process of selecting the appropriate interpretation and generating a response tailored to the student's level of expertise. The Student Model keeps a record of students' performance history which provides information essential to the analysis of multiple parses, multiple errors, and the level of interaction with the student.</Paragraph>
    <Paragraph position="5"> In the German Tutor, the ILTS described, the process leading to the creation of an instructional message in the event of an error has  three stages: (1) Given a forest of parse trees created by the grammar and parser, the parse most likely representative of the intentions of the student must be selected; (2) In the cases when the parse  representing a student's intentions contains several errors, one of the error must be selected as the one that will be addressed. This step is necessary because empirical studies have found that reporting all the errors in a sentence is pedagogically inappropriate. For example, in evaluating her own system Schwind  (1990) reports that &amp;quot;\[s\]ometimes, however, the explanations were too long, especially when students accumulated errors.&amp;quot;~; (3) Given an error, an instructional message must be constructed that is  appropriate to the student's level of expertise and background.</Paragraph>
    <Paragraph position="6"> In Section 1, the theory behind the grammar and its formalism is briefly discussed. Section 2 describes the process leading to the selection of a particular parse and how the Student Model participates in this process. We further discuss the pedagogic role of the Student Model in handling multiple errors and deciding on the level of interaction with the student. Section 3 presents conclusions and Section 4 looks at further research.</Paragraph>
  </Section>
class="xml-element"></Paper>