<?xml version="1.0" standalone="yes"?>
<Paper uid="P06-1108">
  <Title>Event Extraction in a Plot Advice Agent</Title>
  <Section position="10" start_page="862" end_page="863" type="concl">
    <SectionTitle>
6 Conclusion and Future Work
</SectionTitle>
    <Paragraph position="0"> Since the task involved a fine-grained analysis of the rewritten story, the use of events that take plot structure into account made sense regardless of its performance. The use of events as structured features in a machine-learning classifier outperformed a classifier that relied on a unstructured &amp;quot;bag-of-words&amp;quot; as features. The system achieved close to human performance on rating the stories.</Paragraph>
    <Paragraph position="1"> Since each of the events used as a feature in the machine-learner corresponds to a particular event in the story, the features are easily interpretable by other components in the system and interpretable by humans. This allows these events to be used in a template-driven system to generate advice for students based on the structure of their plot.</Paragraph>
    <Paragraph position="2"> Extracting events from text is fraught with error, particularly in the ungrammatical and informal domain used in this experiment. This is often a failure of our system to detect semantic content units through either not including them in chunks or only partially including a single unit in a chunk.</Paragraph>
    <Paragraph position="3"> Chunking also has difficulty dealing with prepositions, embedded speech, semantic role labels, and complex sentences correctly. Improvement in our ability to retrieve semantics would help both story classification and advice generation.</Paragraph>
    <Paragraph position="4"> Advice generation was impaired by the ability to produce directed questions from the events using templates. This is because while our system could detect important events and their or- null der, it could not make explicit their connection through inference. Given the lack of a large-scale open-source accessible &amp;quot;common-sense&amp;quot; knowledge base and the difficulty in extracting inferential statements from raw text, further progress using inference will be difficult. Progress in either making it easier for a teacher to make explicit the important inferences in the text or improved methodology to learn inferential knowledge from the text would allow further progress. Tantalizingly, this ability for a reader to use &amp;quot;inference to grasp points that may not have been explicit in the story&amp;quot; is given as the hallmark of truly understanding a story by teachers.</Paragraph>
  </Section>
class="xml-element"></Paper>