<?xml version="1.0" standalone="yes"?>
<Paper uid="P06-2024">
  <Title>Sydney, July 2006. c(c)2006 Association for Computational Linguistics Towards A Modular Data Model For Multi-Layer Annotated Corpora</Title>
  <Section position="3" start_page="0" end_page="0" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> Five approaches to representing multi-layer annotated corpora are reviewed in this paper. These reflect the current practice in the field and show the requirements typically posed on multi-layer corpus applications. Multi-layer annotated corpora keep annotations at different levels of linguistic organization separate from each other. Figure 1 illustrates two annotation layers on a transcription of an audio/video signal. One layer contains a functional annotation of a sentence in the transcription. The other contains a phrase structure annotation and Part-of-Speech tags for each word.</Paragraph>
    <Paragraph position="1"> Layers and signals are coordinated by a common timeline.</Paragraph>
    <Paragraph position="2"> The motivation for this research is rooted in finding a proper data model for PACE-Ling (Sec. 2.2). The ultimate goal of our research is to create a modular extensible data model for multi-layer annotated corpora. To achieve this, we aim to create a data model based on the current state-of-the-art that covers all current requirements and  base data then decompose it into exchangeable components.</Paragraph>
    <Paragraph position="3"> We identify and discuss objects contained in four  tierscommonlyplayinganimportantroleinmultilayer corpus scenarios (see Fig. 2): medial, locational, structural and featural tiers. Thesearegeneralized categories that are in principle present in any multi-layer context, but come in different incarnations. Since query language and data model are closely related, common query requirements are also surveyed and examined for modular decomposition. While parts of the suggested data modelandqueryoperatorsareimplementedbythe projects discussed here, so far no comprehensive implementation exists.</Paragraph>
  </Section>
class="xml-element"></Paper>