<?xml version="1.0" standalone="yes"?>
<Paper uid="W04-0410">
  <Title>Frozen Sentences of Portuguese: Formal Descriptions for NLP</Title>
  <Section position="2" start_page="0" end_page="1" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> Defining frozen sentences is not easy, and many conceptual and terminological disputes on concurrent terms ('idioms', 'collocations', 'phrasemes', etc.) can be found in the literature (M. Gross 1988; G.Gross 1996; Mejri 1997; Mel'cuk 1993; Mogorron-Huerta 2002; Gaatone 2000; Jurafsky &amp; Martin 2000: 571-573; McKeown &amp; Rodev 2000; Mutsimoto 2003: 395). As a first approach to a (consensual?) definition, frozen sentences are elementary sentences where the main verb and at least one of its argument noun-phrases are distributionally constraint, and usually the global meaning of the expression cannot be calculated from the individual meaning of its component elements when they are used independently (M. Gross 1982, 1989, 1996; G. Gross 1996; Ranchhod 2003). For that reason, the whole expression must be taken as a complex, multiword lexical unit. For example, in: (1) O Joao matou dois coelhos de uma cajadada (lit: 'John killed two rabbits with one blow', John killed two birds with one stone) the verb-object combination (matar-coelhos) is frozen. One cannot replace coelhos (rabbits) for another animal and the numeral determinant is necessarily dois (two).</Paragraph>
    <Paragraph position="1"> Also, it is not possible to modify coelhos with any free adjective (dois coelhos *gordos/ *grandes, two fat/big rabbits). In addition, the adverbial de uma cajadada (with one blow) can hardly be zeroed, or the meaning of the sentence becomes literal. On the other hand, frozen sentences usually present some, often highly constraint, formal variation. For the most part, this variation is strictly lexical. In this case, in the adverbial, the noun cajadada (lit: a blow with a stick) can be replaced by assentada and vez (turn), but the meaning of the expression remains unchanged. This variation does not happens elsewhere. Furthermore, if we disregard proverbs  , completely frozen sentences are rare. Usually, one or more of its argument noun phrases are distributionally free. In this case, any human noun can adequately occupy the structural position of subject. The frozen verb-noun combination is responsible for this distributional constraint, which can be considerably different from the constraints imposed by the verb when functioning as an independent lexical unit. For example, the verb matar (to kill) admits both human and non-human (animate and abstract) nouns for its subject when its object is coelhos (rabbits): (2a) O Joao/a raposa/o tiro/a praga matou dois coelhos (John/the fox/ shooting/ plague killed two rabbits).</Paragraph>
    <Paragraph position="2"> Another way frozen sentences often differ from free sentences is the fact that they block transformations that should otherwise be possible, given the syntactic properties of the main verb and its arguments. Hence, while it is possible to form from (2a) the passive sentence (2b): (2b) Dois coelhos foram mortos pelo Joao/ a raposa/ o tiro/ a praga (Two rabbits were killed by John/ the fox/ shoot- null Proverbs differ from frozen sentences (a) from not having any free elements, (b) in the way they function in discourse, which is similar to quotations and (c) for their general value as advices or as atemporal truths about human life. However, partly because of their being an oral tradition, even proverbs can show some formal variation (Chacoto 1994).</Paragraph>
    <Paragraph position="3"> Second ACL Workshop on Multiword Expressions: Integrating Processing, July 2004, pp. 72-79 ing/ plague), the same does not happen with (1): (1a) oDois coelhos foram mortos pelo Joao de uma cajadada (Two rabbits were killed by John with one blow), since the meaning of the sentence becomes literal (this is shown by symbol 'o'). Finally, frozen sentences constitute a non-trivial problem for many NLP applications. Since they are multi-word expressions functioning as meaning units, they have to be identified as a block. However, their elements can appear discontinuously and they can also present some formal variation. They are often ambiguous, the same sequence having both a literal and a figurative meaning - and in this case, only an extended context can disambiguate them.</Paragraph>
    <Paragraph position="4"> They constitute an important part of the lexicon, comparable to (but probably much larger than) simple verbs.</Paragraph>
  </Section>
class="xml-element"></Paper>