<?xml version="1.0" standalone="yes"?>
<Paper uid="P98-2209">
  <Title>Reactive Content Selection in the Generation of Real-time Soccer Commentary</Title>
  <Section position="3" start_page="0" end_page="1282" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> Timeliness, or reactivity, plays an important role in actual language use. An expression should not only be appropriately planned to communicate relevant content, but should also be uttered at the right moment to describe the action and further to carry on the discourse smoothly. Content selection and its generation are inseparable here. For example, people often start talking before knowing all that they want to say. It is also relatively common' to fill gaps in commentary by describing what was true inthe past. An extreme instance is when an utterance needs to be interrupted to describe a more important event that suddenly occurs.</Paragraph>
    <Paragraph position="1"> It might be expected that dialogue systems should have addressed such real-time issues, but in fact these studies appear to have been much more focused on content planning. The reason for this lies in the nature of dialogue. Although many human-human conversations involve a lot of time pressure, slower conversations can also be successful provided the planning is sufficiently incorporated. For example, even if one conversation participant spends time before taking a turn, the conversation partner can just wait until hearing a contribution.</Paragraph>
    <Paragraph position="2"> In contrast, reactivity is inevitable in live commentary generation, because the complexity and the rapid flow of the situation severely restrict what to be said, and when. If too much time is spent thinking, the situation will unfold quickly into another phase and important events will not be mentioned at the right time.</Paragraph>
    <Paragraph position="3"> MIKE is an automatic narration system that generates spoken live commentary of a simulated soccer game in English, French, or Japanese. We chose the game of soccer firstly because it is a multi-agent game in which various events happen simultaneously in the field. Thus, it is a suitable domain to study real-time content selection among many heterogeneous facts. A second reason for choosing soccer is that detailed, high-quality logs of simulated soccer games are available on a real-time basis from Soccer Server(Noda and Matsubara, 1996), the official soccer simulation system for the RoboCup (Robotic Soccer World Cup) initiative.</Paragraph>
    <Paragraph position="4"> The rest of the paper proceeds as follows. First, we describe our principle for real time content selection and explain its background. Then, after briefly explaining MIKE'S overall design, SS4 explains how our principles are realized within our implementation. SS6 discusses some related works, and SS5 presents some actual output by MIKE and evaluates it in terms of efficiency of communication.</Paragraph>
    <Paragraph position="5">  2 Principles of Content Selection in the Real Time Discourse</Paragraph>
    <Section position="1" start_page="0" end_page="1282" type="sub_section">
      <SectionTitle>
2.1 Maximization of Total Information
</SectionTitle>
      <Paragraph position="0"> A discourse is most effective when the amount of information transmitted to the listener is maximal.</Paragraph>
      <Paragraph position="1"> In the case O f making discourse about a static sub-ject whose situation does not change, the most important contents can be selected and described in  the given time.</Paragraph>
      <Paragraph position="2"> In the case of making discourse on adynamic subject, however, content selection suddenly becomes very complex. Above all, the importance of the contents changes according to the dynamic discourse topic, and also according to the dynamic situation of the subject. Additionally, past events become less importarit with time. Under this condition, the basic function of content selection is to choose the most important content at any given time. This control, however, is not enough, because any content will take time to be uttered and during that time, the situation of the subject might change rapidly.</Paragraph>
      <Paragraph position="3"> Therefore, it should always be possible to change or rearrange the content being uttered.</Paragraph>
      <Paragraph position="4"> Examples of such rearrangements are: * interruption. When the situation of the sub-ject changes suddenly to a new one, more information can be given by rejecting the current utterance and switching to new one.</Paragraph>
      <Paragraph position="5"> * abbreviation. When many important facts arise, the total information can be augmented by referring to each facts quickly by abbreviating each one.</Paragraph>
      <Paragraph position="6"> * repetition. When nothing new comes up in the subject, the important facts already uttered can be repeated to reinforce the information given to the listener.</Paragraph>
      <Paragraph position="7"> As a consequence, creating a system which involves real time discourse concerns 1.assessing the dynamic importance of contents, 2.controlling the content selection with this importance so that the total information becomes maximal using the rearrangement functions.</Paragraph>
      <Paragraph position="8"> In SS4, we discuss how we implemented these principles in MIKE to produce a real time narration.</Paragraph>
    </Section>
    <Section position="2" start_page="1282" end_page="1282" type="sub_section">
      <SectionTitle>
2.2 What, How and When-to-Say
</SectionTitle>
      <Paragraph position="0"> The previous section pointed out that contents should be uttered at the right time; that is, real time discourse systems should effectively address the problem of when-to-say any piece of information.</Paragraph>
      <Paragraph position="1"> However, in MIKE we have only an implicit model of when-to-say. Rather, a collection of game analysis modules and inference rules first suggest the possible comments that can be made (what-to-say). Then, an NL-generation module decides which of these comments to say (again what-to-say), and also how it should be realised (how-to-say). This how-to-say process takes into account issues such as the rearrangements described in the previous section.</Paragraph>
      <Paragraph position="2"> In traditional language generation research, the relationship between the what-to-say aspect (planning) and the how-to-say aspect (surface generation) * Explanation of complex events concern form changes, position change, and advanced plays.</Paragraph>
      <Paragraph position="3"> * Evaluation of team plays concern average forms, forms at a certain moment, players' location, indication of the active or problematic players, winning passwork patterns, wasteful movements.</Paragraph>
      <Paragraph position="4">  has been widely discussed (Appelt, 1982) (Hovy, 1988). One viewpoint is that, for designing natural language systems, it is better to realize what-to-say and how-to-say as separate modules. However, in MIKE we found that the time pressure in the domain makes it difficult to separate what-to-say and how-to-say in this way. Our NL generator decides both on what-to-say and how-to-say because the rearrangements made when deciding how to realize a piece of information directly affect the importance of the remaining unuttered comments. To separate these processes cause significant time delays that would not be tolerable in our time-critical domain.</Paragraph>
    </Section>
  </Section>
class="xml-element"></Paper>