<?xml version="1.0" standalone="yes"?>
<Paper uid="N06-4005">
  <Title>the Semantic Web</Title>
  <Section position="3" start_page="2" end_page="269" type="intro">
    <SectionTitle>
, and
</SectionTitle>
    <Paragraph position="0"> the use of generic lexical resources, such as WordNet.</Paragraph>
    <Paragraph position="1"> Moreover, AquaLog learning mechanism ensures that, for a given ontology and a particular community jargon used by end users, its performance improves over time, as the users can easily correct mistakes and allow AquaLog to learn novel associations between the NL relations used by users and the ontology structure.</Paragraph>
    <Paragraph position="2"> Approach. AquaLog uses a sequential process model (see Fig. 1), in which NL input is first translated into a set of intermediate representations called Query Triples, by the Linguistic Component. The Linguistic Component uses the GATE infrastructure and resources (Cunningham, 2002) to obtain a set of syntactic annotations associated with the input query and to classify the query.</Paragraph>
    <Paragraph position="3"> Once this is done, it becomes straight-forward for the Linguistic Component to automatically create the Query-Triples. Then, these query triples are  A plug-in mechanism and a generic API ensure that different Knowledge Representation languages can be used.</Paragraph>
    <Paragraph position="4">  further processed and interpreted by the Relation Similarity Service Component (RSS), which uses lexical resources and the ontology to map them to ontology-compliant semantic markup or triples.</Paragraph>
  </Section>
class="xml-element"></Paper>