<?xml version="1.0" standalone="yes"?>
<Paper uid="W04-2613">
  <Title>Generating Linear Orders of Text-Based Events</Title>
  <Section position="8" start_page="1" end_page="1" type="concl">
    <SectionTitle>
7 Conclusions and future work
</SectionTitle>
    <Paragraph position="0"> This paper introduces a method for abstracting events from text and generating a linear order of these events based on the temporal semantics associated with them. These simpler orders are useful as summaries of the events contained in the textual description. They also provide a basis for reasoning about the events and performing operations on the orders, such as returning, for example, all events that happened before event X, or returning the first event, last event, etc. The approach presented in the paper models events as event interval relations and maps the thirteen possible event interval relations to before and equals. A system of constraints that preserve the semantics associated with the original event relations assists in generating orders that are plausible. For example, when one event is during another, no intermediate events will fall between these two events in a linear order. Since for many cases not all relations among events are known, it is possible that there are many candidate linear orders produced for a given narrative. In an effort to reduce the number of orders, methods are presented for filtering the orders based on event relations.</Paragraph>
    <Paragraph position="1"> This linear sequencing helps us to understand and communicate how events described in text occur over time and serves as the basis for establishing a time-line of events. Timelines organize events into chronological order and provide a method of encapsulation or summarization where detailed data are rendered in a simpler, temporally ordered fashion.</Paragraph>
    <Paragraph position="2"> Future work will focus on additional approaches for filtering possible orders, for example, if the narrative includes the locations of events, this spatial information can be used to refine further the orders. Events that happen beyond a certain distance, or outside a known area of interest, for example, can be abstracted from the orders.</Paragraph>
  </Section>
class="xml-element"></Paper>