<?xml version="1.0" standalone="yes"?>
<Paper uid="C04-1153">
  <Title>Learning Greek Verb Complements: Addressing the Class Imbalance</Title>
  <Section position="3" start_page="1" end_page="1" type="intro">
    <SectionTitle>
2 Modern Greek
</SectionTitle>
    <Paragraph position="0"> Concerning morphology, MG is highly inflectional. The part-of-speech (pos), the grammatical case, and the verb voice are key morphological features for complement detection.</Paragraph>
    <Paragraph position="1"> Concerning sentence structure, MG is a 'semifree' word-order language. The arguments of a verb do not have fixed positions with respect to the verb and are therefore determined primarily by their morphology rather than their position.</Paragraph>
    <Paragraph position="2"> Certain semantic verb attributes are also very significant: the verb's copularity, its mode, and whether it is (im)personal. A verb is copular when it assigns a quality to its subject. Mode is the prop-erty that determines the semantic relation between the verb and its subject (whether the latter affects or is affected by the verb action. Although all of these features are normally context-dependent, there are verbs with apriori known values for them.</Paragraph>
    <Paragraph position="3"> This apriori information is taken into account in our final dataset, as context-dependent semantic information could not be provided automatically, and we tried to keep manual intervention to a minimum.</Paragraph>
    <Paragraph position="4"> In MG, verbs can take zero, one or two complements. A complement may be a noun phrase in the accusative or the genitive case, a prepositional phrase or a secondary clause (Klairis and Babiniotis 1999). Often the complements appear within the verb phrase itself in the form of weak personal pronouns. Copular verbs only can take as an argument a noun or adjective in the nominative (predicative). Each of the above features is important but not definitive on its own for complement detection.</Paragraph>
    <Paragraph position="5"> When combined, however, and including context information of the candidate complement, many cases of ambiguity are correctly resolved. The biggest sources of ambiguity are the accusative noun phrase, which is very often adverbial denoting usually time, and the prepositional phrase introduced by se (to), also often adverbial, denoting usually place.</Paragraph>
  </Section>
class="xml-element"></Paper>