<?xml version="1.0" standalone="yes"?>
<Paper uid="E06-2026">
  <Title>Grammatical Role Labeling with Integer Linear Programming</Title>
  <Section position="2" start_page="0" end_page="0" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> An often stressed point is that the most widely used classifiers such as Naive Bayes, HMM, and Memory-based Learners are restricted to local decisions only. With grammatical role labeling, for example, there is no way to explicitly express global constraints that, say, the verb&amp;quot;to give&amp;quot;must have 3 arguments of a particular grammatical role.</Paragraph>
    <Paragraph position="1"> Among the approaches to overcome this restriction, i.e. that allow for global, theory based constraints, Integer Linear Programming (ILP) has been applied to NLP (Punyakanok et al., 2004) .</Paragraph>
    <Paragraph position="2"> Weapply ILPto the problem of grammatical relation labeling, i.e. given two chunks.1 (e.g. a verb and a np), what is the grammatical relation between them (if there is any). We have trained a maximum entropy classifier on vectors with morphological, syntactic and positional information.</Paragraph>
    <Paragraph position="3"> Its output is utilized as weights to the ILP component which generates equations to solve the following problem: Given subcategorization frames (expressed in functional roles, e.g. subject), and given a sentence with verbs, a0 (auxiliary, modal, finite, non-finite, ..), and chunks, a1 (a2a4a3 ,a3a5a3 ), label all pairs (a0a7a6 a1a9a8a11a10a7a12 a0a7a6 a1 ) withagrammatical role2. Inthispaper, wearepursuing twoempirical scenarios. The first is to collapse all subcategoriza1Currently, we use perfect chunks, that is, chunks stemming from automatically flattening a treebank.</Paragraph>
    <Paragraph position="4"> 2Most of these pairs do not stand in a proper grammatical relation, they get a null class assignment.</Paragraph>
    <Paragraph position="5"> tion frames of a verb into a single one, comprising all subcategorized roles of the verb but not necessarily forming a valid subcategorization frame of that verb at all. For example, the verb &amp;quot;to believe&amp;quot; subcategorizes for a subject and a prepositional complement (&amp;quot;He believes in magic&amp;quot;) or for a subject and a clausal complement (&amp;quot;She believes that he is dreaming&amp;quot;), but there is no frame that combines a subject, a prepositional object and a clausal object. Nevertheless, the set of valid grammatical roles of a verb can serve as a filter operating upon the output of a statistical classifier. The typical errors being made by classifiers with only local decisions are: a constituent is assigned to a grammatical role more than once and a grammatical role (e.g. of a verb) is instantiated more than once. The worst example in our tests was a verb that receives from the maxent classifier two subjects and three clausal objects. Here, such a role filter will help to improve the results.</Paragraph>
    <Paragraph position="6"> The second setting is to provide ILP with the correct subcategorization frame of the verb. The results of such an oracle setting define the upper bound of the performance our ILP approach can achieve. Future work will be to let ILP find the optimal subcategorization frame given all frames of a verb.</Paragraph>
  </Section>
class="xml-element"></Paper>