<?xml version="1.0" standalone="yes"?>
<Paper uid="C02-2025">
  <Title>The LinGO Redwoods Treebank Motivation and Preliminary Applications</Title>
  <Section position="4" start_page="0" end_page="0" type="relat">
    <SectionTitle>
4 Related Work
</SectionTitle>
    <Paragraph position="0"> To the best of our knowledge, no prior research has been conducted exploring the linguistic depth, flexibility in available information, and dynamic nature of tree-banks that we have proposed. Earlier work on building corpora of hand-selected analyses relative to an existing broad-coverage grammar was carried out at Xerox PARC, SRI Cambridge, and Microsoft Research. As all these resources are tuned to proprietary grammars and analysis engines, the resulting treebanks are not publicly available, nor have reported research results been reproducible. Yet, especially in light of the successful LinGO open-source repository, it seems vital that both the tree-bank and associated processing schemes and stochastic models be available to the general (academic) public. An on-going initiative at Rijksuniversiteit Groningen (NL) is developing a treebank of dependency structures (Mullen, Malouf, &amp; Noord, 2001), derived from an HPSG-like grammar of Dutch (Bouma, Noord, &amp; Malouf, 2001).</Paragraph>
    <Paragraph position="1"> The general approach resembles the Redwoods initiative (specifically the discriminator-based method of tree selection; the LKB tree comparison tool was originally developed by Malouf, after all), but it provides only a single stratum of representation, and has no provision for evolving analyses in tandem with the grammar. Dipper (2000) presents the application of a broad-coverage LFG grammar for German to constructing tectogrammatical structures for the TiGer corpus. The approach is similar to the Groningen framework, and shares its limitations.</Paragraph>
  </Section>
class="xml-element"></Paper>