<?xml version="1.0" standalone="yes"?>
<Paper uid="W04-1707">
  <Title>Towards Deeper Understanding and Personalisation in CALL</Title>
  <Section position="2" start_page="0" end_page="0" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> The tendency to develop natural interfaces for all users implies man-machine interaction in a natural way, including natural language too, both as speech and as free text. Many recent eLearning research prototypes try to cope with the unrestricted text input as it is considered old-fashioned and even obsolete to offer interfaces based on menu-buttons and mouse-clicking communication only. On the other hand, the available eLearning platforms such as WebCT [1], CISCO [2], and the freeware HotPotatoes [3], are far from the application of advanced language technologies that might provide interfaces based on speech and language processing. They represent complex communication environments and/or empty shells where the teacher uploads training materials, drills, etc. using specialised authoring tools. Recently on-line voice communication between teachers and students has been made available as well, via fast Internet in virtual classrooms, but no speech or language processing has been considered. So there is a deep, principle gap between the advanced research on tutoring systems and the typical market eLearning environments addressing primarily the communication needs of the mass user.</Paragraph>
    <Paragraph position="1"> In what follows we will concentrate on research prototypes integrating language technologies in eLearning environments. In general, such prototypes might be called Intelligent Tutoring Systems (ITS) and we will stick to this notion here. Most of the systems discussed below address Computer-Aided Language Learning (CALL) but language technologies are applied for automatic analysis of user utterances in other domains too. A review of forty Intelligent CALL systems (Gamper, 2002) summarises the current trends to embed &amp;quot;intelligence&amp;quot; in CALL. What we developed (and report here) might be considered intelligent because of the integration of reasoning and the orientation to adaptivity and personalisation.</Paragraph>
    <Paragraph position="2"> This paper is structured as follows. In section 2 we consider the task of semantic analysis of the learner's input, which is an obligatory element when the student is given the opportunity to type in freely in response to ITS's questions and/or drills. Section 3 deals with Information Retrieval (IR) approaches for measuring document similarity, which are integrated in ITS as techniques for e.g.</Paragraph>
    <Paragraph position="3"> assessing the content of student essays or choosing the most relevant text to be shown to the learner.</Paragraph>
    <Paragraph position="4"> Section 4 discusses how the language technologies in question can provide some adaptivity of the ITS, as a step towards personalisation. In section 5 we summarise the current results regarding the evaluation of our prototypes with real users.</Paragraph>
    <Paragraph position="5"> Section 6 contains the conclusion.</Paragraph>
  </Section>
class="xml-element"></Paper>