<?xml version="1.0" standalone="yes"?>
<Paper uid="W04-2417">
  <Title>A transformation-based approach to argument labeling</Title>
  <Section position="4" start_page="0" end_page="0" type="metho">
    <SectionTitle>
3 Task Definition
</SectionTitle>
    <Paragraph position="0"> Defining the task of semantic role labeling in TBL terms requires four basic steps. First, the problem has to be reduced to that of assigning an appropriate tag to each word in a sentence. Second, we must define the features associated with each word in the sentence, on which the transformational rules will operate. Third, we must decide on the exact forms the transformational rules will be allowed to take (the rule templates). Finally, we must determine a mapping from our word-by-word tag assignment to the labeled bracketing used to identify semantic arguments in the test data. Each of these steps is addressed below.</Paragraph>
    <Section position="1" start_page="0" end_page="0" type="sub_section">
      <SectionTitle>
3.1 Tagging scheme
</SectionTitle>
      <Paragraph position="0"> The simplest way of representing the chunks of text which correspond to semantic arguments is to use some variant of the IOB tagging scheme (Sang and Veenstra, 1999). This is the approach taken by Hacioglu et al. (2003), who apply the IOB2 tagging scheme in their word-by-word models, as shown in the second row of Figure 1.</Paragraph>
      <Paragraph position="1"> However, two aspects of the problem at hand make this tag assignment difficult to use for TBL. First, semantic argument chunks can be very large in size. An argument which contains a relative clause, for example, can easily be longer than 20 words. Second, the label an argument is assigned is largely arbitrary, in the sense that core argument labels (A0, A1, etc.) generally cannot be assigned without some information external to the constituent, such as the class of the predicate, or the identity of other arguments which have already been assigned. So using the IOB2 format, it might take a complicated sequence of TBL rules to completely re-tag, say, an A0 argument as A1. If this re-tagging is imperfectly achieved, we are left with the difficult decision of how to interpret the stranded I-A0 elements, and the problem that they may incorrectly serve as an environment for other transformational rules.</Paragraph>
      <Paragraph position="2"> For this reason, we adopt a modified version of the IOB2 scheme which is a compromise between addressing the tasks of argument identification and argument labeling. The left boundary (B) tags indicate the label of the argument, but the internal (I) tags are non-specific as to argument label, as in the last row of Figure 1. This allows a a single TBL rule to re-label an argument, while still allowing for interleaving of TBL rules which affect argument identification and labeling.</Paragraph>
    </Section>
    <Section position="2" start_page="0" end_page="0" type="sub_section">
      <SectionTitle>
3.2 Feature Coding
</SectionTitle>
      <Paragraph position="0"> With each word in a sentence, we associate the following features: Word The word itself, normalized to lower-case. Tag The word's part-of-speech tag, as predicted by the system of Gim'enez and M`arquez (2003).</Paragraph>
      <Paragraph position="1"> Chunk The chunk label of the word, as predicted by the system of Carreras and M`arquez (2003).</Paragraph>
      <Paragraph position="2"> Entity The named-entity label of the word, as predicted by the system of Chieu and Ng (2003).</Paragraph>
      <Paragraph position="3"> L/R A feature indicating whether the word is to the left (L) or right (R) of the target verb.</Paragraph>
      <Paragraph position="4"> Indent This feature indicates the clause level of the current word with respect to the target predicate. Using the clause boundaries predicted by the system of Carreras and M`arquez (2003), we compute a feature based on the linguistic notion of c-command.1 If both the predicate and the current word are in the same basic clause, Indent=0. If the predicate c-commands the current word, and the current word is one clause level lower, Indent=1. If it is two clause levels lower, Indent=2, and so on. If the c-command relations are reversed, the indent levels are negative, and if neither c-commands the other, Indent='NA'.</Paragraph>
      <Paragraph position="5"> (Figure 2 illustrates how this feature is defined.) The absolute value of the Indent feature is not permitted to exceed 5.</Paragraph>
      <Paragraph position="6"> is-PP A boolean feature indicating whether the word is included within a base prepositional phrase. This is 1A node (reflexively) c-commands a node iff there is a node such that directly dominates , and dominates .</Paragraph>
      <Paragraph position="7"> Note that only clauses (S nodes) are considered in our application described above.</Paragraph>
      <Paragraph position="8"> true if its chunk tag is B-PP or I-PP, or if it is within an NP chunk directly following a PP chunk.</Paragraph>
      <Paragraph position="9"> PP-head If is-PP is true, this is the head of the prepositional phrase; otherwise it is zero.</Paragraph>
      <Paragraph position="10"> N-head The final nominal element of the next NP chunk at the same indent level as the current word, if it exists. For purposes of this feature, a possessive NP chunk is combined with the following NP chunk.</Paragraph>
      <Paragraph position="11"> Verb The target predicate under consideration.</Paragraph>
      <Paragraph position="12"> V-Tag The POS tag of the target predicate.</Paragraph>
      <Paragraph position="13"> V-Passive A boolean feature indicating whether the target verb is in the passive voice. This is determined using a simple regular expression over the sentence.</Paragraph>
      <Paragraph position="14"> Path As in (Pradhan et al., 2003), this feature is an ordered list of the chunk types intervening between the target verb and the current word, with consecutive NP chunks treated as one.</Paragraph>
    </Section>
    <Section position="3" start_page="0" end_page="0" type="sub_section">
      <SectionTitle>
3.3 Rule Templates
</SectionTitle>
      <Paragraph position="0"> In order to define the space of rules searched by the TBL algorithm, we must specify a set of rule templates, which determine the form transformational rules may take. The rule templates used in our system are 130 in number, and fall into a small number of classes, as described below.</Paragraph>
      <Paragraph position="1"> These rules all take the form f1 : : : fn ! labelw, where f1 through fn are features of the current word w or words in its environment, and usually include the current (semantic argument) label assigned to w. The categorization of rule templates below, then, basically amounts to a list of the different feature sets which are used to predict the argument label of each word.</Paragraph>
      <Paragraph position="2"> The initial assignment of tags which is given to the TBL algorithm is a very simple chunk-based assignment.</Paragraph>
      <Paragraph position="3"> Every word is given the tag O (outside all semantic arguments), except if it is within an NP chunk at Indent level zero. In that case, the word is assigned the tag I if its chunk label is I-NP, B-A0 if its chunk label is B-NP and it is to the left of the verb, and B-A1 if its chunk label is B-NP and it is to the right of the verb.</Paragraph>
      <Paragraph position="4"> 3.3.1 Basic rules (10 total) The simplest class of rules simply change the current word's argument label based on its own local features, including the current label, and the features L/R, Indent, and Chunk.</Paragraph>
      <Paragraph position="5">  An expanded set of rules using all features of the current word, as well as the argument labels of the current and previous words. For example, the following rule will change the label O to I within an NP chunk, if the initial</Paragraph>
      <Paragraph position="7"/>
      <Paragraph position="9"> These rules change the argument label of the current word based on the Word feature of the current or surrounding words, in combination with argument labels and chunk labels from the surrounding context. For example, this rule marks the adverb back as a directional modifier when it follows the target verb:</Paragraph>
      <Paragraph position="11"> These rules further add the named-entity tag of the current, preceding, or following word to the basic and local-context rules above.</Paragraph>
      <Paragraph position="12"> 3.3.5 Verb tag (15) These rules add the POS tag of the predicate to the basic and simpler local-context rules above.</Paragraph>
      <Paragraph position="13">  These rules allow the argument label of the current word to be changed, based on its Verb and N-head features,as well as other local features.</Paragraph>
      <Paragraph position="14">  These rules allow the argument label of the current word to be changed, based on its Word, N-head, Indent, L/R, and Chunk features, as well as the argument labels of adjacent words.</Paragraph>
      <Paragraph position="15">  Because many of the dependencies involved in the semantic role labeling task hold over the domain of the entire sentence, we include a number of long-distance rules. These rules allow the argument label to be changed depending on the word's current label, the features L/R, Indent, Verb, and the argument label of a word within 50 or 100 words of the current word. These rules are intended to support generalizations like &amp;quot;if the current word is labeled A0, but there is already an A0 further to the left, change it to I&amp;quot;.</Paragraph>
      <Paragraph position="16"> 3.3.9 &amp;quot;Smoothing&amp;quot; rules (15) Finally, there are a number of &amp;quot;smoothing&amp;quot; rules, which are designed primarily to prevent I tags from becoming stranded, so that arguments which contain a large number of words can successfully be identified. These rules allow the argument label of a word to be changed based on the argument labels of the previous two words, the next two words, and the chunk tags of these words. This sample rule marks a word as being argumentinternal, if both its neighbors are already so marked:</Paragraph>
      <Paragraph position="18"> Finally, we include a number of rule templates using the highly-specific Path feature. These rules allow the argument label of a word to be changed based on its current value, as well as the value of the feature Path in combination with L/R, Indent, V-Tag, Verb, and Word.</Paragraph>
    </Section>
    <Section position="4" start_page="0" end_page="0" type="sub_section">
      <SectionTitle>
3.4 Tag interpretation
</SectionTitle>
      <Paragraph position="0"> The final step in our transformation-based approach to semantic role labeling is to map the word-by word IOB tags predicted by the TBL model back to the format of the original data set, which marks only argument boundaries, so that we can calculate precision and recall statistics for each argument type. The simplest method of performing this mapping is to consider an argument as consisting of an initial labeled boundary tag (such as B-A0, followed by zero or more argument-internal (I) tags, ignoring anything which does not conform to this structure (in particular, strings of Is with no initial boundary marker).</Paragraph>
      <Paragraph position="1"> In fact, this method works quite well, and it is used for the results reported below.</Paragraph>
      <Paragraph position="2"> Finally, there is a post-processing step in which adjucts may be re-labeled if the same sequence of words is found as an adjunct in the training data, and always bears the same role. This affected fewer than twenty labels on the development data, and added only about 0:1 to the overall f-measure.</Paragraph>
    </Section>
  </Section>
class="xml-element"></Paper>