<?xml version="1.0" standalone="yes"?>
<Paper uid="W99-0631">
  <Title>An Iterative Approach to Estimating Frequencies over a Semantic Hierarchy</Title>
  <Section position="3" start_page="0" end_page="258" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> Knowledge of the constraints a verb places on the semantic types of its arguments (variously called selectional restrictions, selectional preferences, selectional constraints) is of use in many areas of natural language processing, particularly structural disambiguation. Recent treatments of selectional restrictions have been probabilistic in nature (Resnik, 1993), (Li and Abe, 1998), (Ribas, 1995), (McCarthy, 1997), and estimation of the relevant probabilities has required corpus-based counts of the number of times word senses, or concepts, appear in the different argument positions of verbs. A difficulty arises due to the absence of a large volume of sense disambiguated data, as the counts have to be estimated from the nouns which appear in the corpus, most of which will have more than one sense. The techniques in Resnik (1993), Li and Abe (1998) and Ribas (1995) simply distribute the count equally among the alternative senses of a noun. Abney and Light (1998) have attempted to obtain selectional preferences using the Expectation Maximization algorithm by encoding WordNet as a hidden Markov model and using a modified form of the forward-backward algorithm to estimate the parameters.</Paragraph>
    <Paragraph position="1"> The approach proposed in this paper is to use a re-estimation process which relies on counts being passed up a semantic hierarchy, from the senses of nouns appearing in the data. We make use of the semantic hierarchy in WordNet (Fellbaum, 1998), which consists of word senses, or concepts, 1 related by the 'is-a' or 'is-a-kind-of' relation. If c' is a kind of c, then c is a hypernym of c', and c' a hyponym of c. Counts for any concept are transmitted up the hierarchy to all of the concept's hypernyms. Thus if eat chicken appears in the corpus, the count is transmitted up to &lt;meat &gt;, &lt; :food&gt;, and all the other hypernyms of that sense of chicken? The problem is how to distinguish the correct sense of chicken in this case from incorrect senses such as &lt;wimp&gt;. 3 We utilise the 1We use the words sense and concept interchangeably to refer to a node in the semantic hierarchy. eWe use italics when referring to words, and angled brackets when referring to concepts or senses. This notation does not always pick out a concept uniquely, but the particular concept being referred to should be clear from the context.</Paragraph>
    <Paragraph position="2"> 3The example used here is adapted from Mc-Carthy (1997). There are in fact four senses of chicken in WordNet 1.6, but for ease of exposition we consider only two. The hypernyms of the  fact that whilst splitting the count equally can lead to inaccurate estimates, counts do tend to accumulate in the right places. Thus counts will appear under &lt;:food&gt;, for the object of eat, but not under &lt;person&gt;, indicating that the object position of eat is more strongly associated with the set of concepts dominated by &lt;:food&gt; than with the set of concepts dominated by &lt; person &gt;.</Paragraph>
    <Paragraph position="3"> By choosing a hypernym for each alternative sense of chicken and comparing how strongly the sets dominated by these hypernyms associate with eat, we can give more count in subsequent iterations to the food sense of chicken than to the wimp sense.</Paragraph>
    <Paragraph position="4"> A problem arises because these two senses of chicken each have a number of hypernyms, so which two should be compared? The chosen hypernyms have to be high enough in the hierarchy for adequate counts to have accumulated, but not so high that the alternative senses cannot be distinguished. For example, a hypernym of the food sense of chicken is &lt;poultry&gt;, and a hypernym of the wimp sense is &lt;weakling&gt;. However, these concepts may not be high enough in the hierarchy for the accumulated counts to indicate that eat is much more strongly associated with the set of concepts dominated by &lt;poultry&gt; than with the set dominated by &lt;weakling&gt;. At the other extreme, if we were to choose &lt;entity&gt;, which is high in the hierarchy, as the hypernym of both senses, then clearly we would have no way of distinguishing between the two senses.</Paragraph>
    <Paragraph position="5"> We have developed a technique, using a X 2 test, for choosing a suitable hypernym for each alternative sense. The technique is based on the observation that a chosen hypernym is too high in the hierarchy if the set consisting of the children of the hypernym is not sufficiently homogeneous with respect to the given verb and argument position. Using the previous example, &lt;entity&gt; is too high to represent either sense of chicken because food sense are &lt;poultry&gt;, &lt;bird&gt;, &lt;meat&gt;, &lt; foodstuff &gt;, &lt; food &gt;, &lt; substance &gt;, &lt; object &gt;, &lt; entity &gt;. The hypernyms of the wimp sense are &lt; weakling &gt;, &lt; person &gt;, &lt; life_form&gt;, &lt;entity&gt;.</Paragraph>
    <Paragraph position="6"> the children of &lt;entity&gt; are not all associated in the same way with eat. The set consisting of the children of &lt;meat&gt;, however, is homogeneous with respect to the object position of eat, and so &lt;meat&gt; is not too high a level of representation. The measure of homogeneity we use is detailed in Section 5.</Paragraph>
  </Section>
class="xml-element"></Paper>