<?xml version="1.0" standalone="yes"?>
<Paper uid="E06-2020">
  <Title>Generating Spatio-Temporal Descriptions in Pollen Forecasts</Title>
  <Section position="2" start_page="0" end_page="0" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> New monitoring devices such as remote sensing systems are generating vast amounts of spatio-temporal data. These devices, coupled with the wider accessibility of the data, have spurred large amounts of research into how it can best be analysed. There has been less research however, into how the results of the data analysis can be effectively communicated. As part of a wider research project aiming to produce textual reports of complex spatio-temporal data, we have developed a prototype NLG system which produces textual pollen forecasts for the general public.</Paragraph>
    <Paragraph position="1"> Pollen forecast texts describe predicted pollen concentration values for different regions of a country.</Paragraph>
    <Paragraph position="2"> Their production involves two subtasks; predicting pollen concentration values for different regions of a country, and describing these numerical values textually.In our work, we focus on the later subtask, textual description of spatio-temporally distributed pollen concentration values. The subtask of predicting pollen concentrations is carried out by our industrial collaborator, Aerospace and Marine International (UK) Ltd (AMI).</Paragraph>
    <Paragraph position="3"> A fairly substantial amount of work already exists on weather forecast generation. A number of systems have been developed and are currently in commercial use with two of the most notable being FOG (Goldberg et al., 1994) and MultiMeteo (Coch, 1998).</Paragraph>
  </Section>
class="xml-element"></Paper>