<?xml version="1.0" standalone="yes"?>
<Paper uid="W97-1015">
  <Title>A Comparative Study of the Application of Different Learning Techniques to Natural Language Interfaces</Title>
  <Section position="2" start_page="0" end_page="0" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> One of the main obstacles to the efficient use of natural language interfaces is the often required high amount of manual knowledge engineering (see (Androutsopoulos et al., 1995) for a recent survey). This time-consuming and tedious process is often referred to as &amp;quot;knowledge acquisition bottleneck&amp;quot;. It may require extensive efforts by experts highly experienced in linguistics as well as in the domain and the task (Riloff and Lehnert, 1994). Therefore, natural language interfaces represent a domain that is very well suited for the application of machine learning algorithms to automate the acquisition process of linguistic knowledge.</Paragraph>
    <Paragraph position="1"> So far, inductive learning has already been applied successfully to a large variety of natural Janguage tasks. This includes basic linguistic problems such as morphological analysis (van den Bosch et al., 1996), parsing (Zelle and Mooney, 1996), word sense disambiguation (Mooney, 1996), and anaphora resolution (Aone and Bennett, 1996). Besides this, there also exists some research on applications, e.g.</Paragraph>
    <Paragraph position="2"> machine translation (Yamazaki et al., 1996), text categorization (Moulinier and Ganaseia, 1996), or information extraction (Soderland et al., 1996).</Paragraph>
    <Paragraph position="3"> The learning task in natural language interfaces is to select the correct command class based on semantic features extracted from the user input. Therefore, it can be modeled as classification problem, i.e. the machine learning algorithms construct a theory from the training data that is used for classifying unseen test data (Quinlan, 1996). So far, we consider only supervised learning so that each training case has to be labeled with the correct class.</Paragraph>
    <Paragraph position="4"> We apply different existing instance-based and model-based algorithms to this problem and compare the achieved results. In addition, we have also developed several new algorithms, which we present briefly in this paper. We have implemented all algorithms by means of the deductive object-oriented database system ROCK ~ ROLL (Barja et al., 1994).</Paragraph>
    <Paragraph position="5"> It solves the problem of updates in deductive databases in that it separates the declarative logic query language ROLL from the imperative data manipulation language ROCK within the context of a common object-oriented data model. Besides this, ROCK ~ ROLL makes a clean distinction between type declarations, which describe the structural characteristics of a set of instance objects and the methods that can be applied to them, and class definitions, which specify the implementation of the methods associated with a type.</Paragraph>
    <Paragraph position="6"> The use of the available powerful logic and object-oriented programming language enables an efficient implementation of the different approaches to machine learning. It also gives us a convenient integrated tool that assists in applying the machine learning algorithms to the data collection stored in the same database.</Paragraph>
    <Paragraph position="7"> Winiwarter ~ Kambayashi 125 Learning and NL Interfaces Werner Winiwarter and Yahiko Kambayashi (1997) A Comparative Study of the Application of Different Learning Techniques to Natural Language Interfaces. In T.M. Ellison (ed.) CoNLL97: Computational Natural Language Learning, ACL pp 125-135. (~) 1997 Association for Computational Lingtfistics</Paragraph>
    <Section position="1" start_page="0" end_page="0" type="sub_section">
      <SectionTitle>
User input
</SectionTitle>
      <Paragraph position="0"> and lexical analysis and lexical analysis ~ I.o. d l exical analysi____,_ _ s  As comparative evaluation of the implemented algorithms, we applied them to an extensive case study: a natural language interface for a production planning and control system. The system is used in a multilingual environment, which includes the languages English, German, and Japanese. Therefore, an important issue of the evaluation was to check whether the learned knowledge is languageindependent, i.e. if it really operates based on semantic deep forms so that it abstracts from linguistic surface phenomena.</Paragraph>
      <Paragraph position="1"> The rest of the paper is organized as follows.</Paragraph>
      <Paragraph position="2"> First, we briefly introduce the learning task before we present the applied machine learning algorithms in more detail. Finally, we explain the set-up of the case study and discuss the achieved results from evaluation.</Paragraph>
    </Section>
  </Section>
class="xml-element"></Paper>