<?xml version="1.0" standalone="yes"?>
<Paper uid="W03-0907">
  <Title>Story understanding through multi-representation model construction</Title>
  <Section position="3" start_page="0" end_page="0" type="metho">
    <SectionTitle>
3 A multi-representation axiomatization
</SectionTitle>
    <Paragraph position="0"> for the Snowman story We have created a multi-representation axiomatization of the commonsense knowledge necessary to understand the Snowman story using Shanahan's event calculus. Table 2 shows how our axiomatization compares with other event calculus axiomatizations.</Paragraph>
    <Paragraph position="1">  man story: It does not deal with the snowman disappearing and melting and the resulting thoughts of James. It does not deal with the creation and destruction of objects such as the snowballs. These are assumed to exist over all time points. CTime: clock time.</Paragraph>
    <Paragraph position="2"> ECTime: the event calculus model of time.</Paragraph>
    <Paragraph position="3"> Feeling: simple positive, neutral, and negative emotions, and positive, neutral, and negative attitudes toward objects. OMSpace: object-scale metric space, with falling and collisions.</Paragraph>
    <Paragraph position="4"> OTSpace: object-scale topological space.</Paragraph>
    <Paragraph position="5"> PlayNeed: the need to play, with a simple model of needs and intentions.</Paragraph>
    <Paragraph position="6"> RTSpace: room-scale topological space.</Paragraph>
    <Paragraph position="7"> Sleep: sleeping and body posture.</Paragraph>
    <Paragraph position="8"> Snow: snow and snow falling from the sky.</Paragraph>
    <Paragraph position="9"> SpeechAct: some simple speech acts.</Paragraph>
    <Paragraph position="10"> Vision: some simple aspects of vision.</Paragraph>
    <Paragraph position="11"> Due to space limitations, in this paper we present only one of the eleven representations: RTSpace.</Paragraph>
    <Section position="1" start_page="0" end_page="0" type="sub_section">
      <SectionTitle>
3.1 Room-scale topological space
</SectionTitle>
      <Paragraph position="0"> The predicates, functions, fluents, and events of RTSpace are shown in Table 3.</Paragraph>
      <Paragraph position="2"> WalkDownStaircase(actor, staircase) actor walks down staircase.</Paragraph>
      <Paragraph position="3"> WalkThroughDoor12(actor, door) actor walks through side one of door.</Paragraph>
      <Paragraph position="4"> WalkThroughDoor21(actor, door) actor walks through side two of door.</Paragraph>
      <Paragraph position="5"> WalkUpStaircase(actor, staircase) actor walks up staircase.</Paragraph>
      <Paragraph position="6">  This representation of space consists of locations (rooms and outside areas), which are connected by portals (doors and staircases).</Paragraph>
      <Paragraph position="7"> A state constraint says that an object is at one location</Paragraph>
      <Paragraph position="9"> A state constraint says that an object is near a portal if and only if there is a location such that the object is at the location and one of the sides of the portal is the location:  We have similar precondition and effect axioms for locking a door.</Paragraph>
      <Paragraph position="10"> A state constraint says that if a door is open, it is unlocked: null Axiom 5.</Paragraph>
      <Paragraph position="11"> HoldsAt(DoorIsOpen(door), time)) HoldsAt(DoorUnlocked(door), time) A precondition axiom states that for an actor to open a door, the actor must be awake, the door must not already be open, the door must be unlocked, and the actor must be near the door: Axiom 6.</Paragraph>
      <Paragraph position="12"> Happens(DoorOpen(actor, door), time)) HoldsAt(Awake(actor), time)^ :HoldsAt(DoorIsOpen(door), time)^ HoldsAt(DoorUnlocked(door), time)^ HoldsAt(NearPortal(actor, door), time) An effect axiom states that if an actor opens a door, the door will be open: Axiom 7.</Paragraph>
      <Paragraph position="13"> Initiates(DoorOpen(actor, door), DoorIsOpen(door), time) We have similar precondition and effect axioms for closing a door.</Paragraph>
      <Paragraph position="14"> Precondition axioms state that for an actor to walk through a side of a door, the actor must be awake and standing, the door must be open, and the actor must be at the side of the door that the actor walks through:  We have similar precondition and effect axioms for walking up and down a staircase.</Paragraph>
      <Paragraph position="15"> A state constraint says that if an actor is outside, the actor is dressed: Axiom 14.</Paragraph>
      <Paragraph position="16"> HoldsAt(At(actor, outside), time)) HoldsAt(Dressed(actor), time) Two locations are adjacent if and only if they have a portal in common:</Paragraph>
    </Section>
  </Section>
  <Section position="4" start_page="0" end_page="2" type="metho">
    <SectionTitle>
4 Processing the Snowman story
</SectionTitle>
    <Paragraph position="0"> The complete run of the Snowman story takes 45 minutes on a machine with a 700 MHz Pentium III processor and 512 megabytes of RAM. Statistics on processing the segments are shown in Table 4.</Paragraph>
    <Paragraph position="1"> Due to space limitations, we cannot show the model of all the story segments. We present here the model of the  Event occurrences are shown at the end of each time point. Only changes in what fluents hold from one time point to the next are shown. Thus after the WakeUp event occurs above, James is no longer asleep and he is awake:  An axiom in the Feeling representation triggers this event in response to the snow: Happens(BecomeHappy(James), 2) An axiom in the PlayNeed representation triggers this event in response to the snow:</Paragraph>
  </Section>
class="xml-element"></Paper>