<?xml version="1.0" standalone="yes"?>
<Paper uid="W97-0108">
  <Title>J Domain-Specific Semantic Class Disambiguation Using WordNet</Title>
  <Section position="3" start_page="56" end_page="57" type="metho">
    <SectionTitle>
2 Our Approach
</SectionTitle>
    <Paragraph position="0"> As opposed to proponents of &amp;quot;domain-specific information for domain-specific applications&amp;quot;, our approach veztures towards the application of general-purpose algor\]t~,~ and resources to our dom~i,specific s~rn~tic class disaznbiguation problem.</Paragraph>
    <Paragraph position="1"> Our information source is the extensive semantic hierarchy WordNet (Miller, 1990) which was designed to capture the semantics of general nuances and uses of the English language. Our approach reconciles the domain-specific hierarchy with this ~ast network and exploits WordNet to uncover semantic ci~s~es, without the need of an ~otated corpus.</Paragraph>
    <Paragraph position="2"> Firstly, the domain-specific hierarchy is mapped onto the semantic network of WordNet, by manually as.~zni~g corresponding WordNet node(s) to the classes in the do~,~-speci~c hierarchy. To disembiguate a word, the sentence context of the word is first streamed through a general word sense disambiguation module which assigns the appropriate sense of the word. The word sense disambiguation module hence effectively pinpoints a partic~l~r node in WordNet that corresponds to the current sense of the word. Thereafter, this chosen concept node is piped througJa a semantic distance module which determines the s~m~c distances between this concept node and all the s~m~,~tic class nodes in the domain-speci~c hierarchy. If the distance between the concept node and a semantic class node is below some threshold, the semantic class node becomes a candidate class node. The nearest candidate eJ~ss node is then chosen as the semautic class of the word.</Paragraph>
    <Paragraph position="3"> If no such candidates exist, the word does not belong to any of the semantic classes in the hierarchy, and is usually labelled as the &amp;quot;entity&amp;quot; class. The flow of our approach is illustrated in Figure 1.</Paragraph>
    <Paragraph position="4"> A walkthrough of the approach with a simple example w~l better illustrate it. Consider a domainspecit~c hierarchy with just 3 classes :- VEHICLE, AIRCRAFT and CAR, as shown in Figure 2(a).</Paragraph>
    <Paragraph position="5"> Mapping this domainospeci~c hierarchy to Word-Net simply involves finding the specific sense(s) of  m r motor_vehicle: 1 Figure 2 : (a) A simple domain-specific hierarchy (b) The classes of the domain-specific hierarchy as mapped onto WordNet, together with the word to be dis~mhigtmted, &amp;quot;plane'.</Paragraph>
    <Paragraph position="6"> the classes. In this case, all three classes correspond to their first sense in WordNet.</Paragraph>
    <Paragraph position="7"> Then, given a sentence, say, &amp;quot;The plane win be taking off in 5 minutes time. ~, to dis~m~iguate the semantic class of the word &amp;quot;plane&amp;quot;, the sentence is fed to the word sense disambiguation module. The module win determine the sense of this wor&amp; In this example, the correct sense of &amp;quot;plane&amp;quot; is sense 1, i.e. the sense of an aeroplane. Having identified the particular concept node in Word.Net that &amp;quot;plaue&amp;quot; corresponds to, the distances between this concept node and the three semantic class nodes are then calculated by the semantic distance module. Based on WordNet, the module will conclude that the concept node &amp;quot;plane:l&amp;quot; is nearer to the semantic class node &amp;quot;aircraft:l&amp;quot; and should hence be cl~Lssified as AIR-CRAFT. Figure 2(b) shows the relative positions of the concept node ~plane:l ~ and the three semantic cl~q nodes in Word_Net.</Paragraph>
    <Section position="1" start_page="57" end_page="57" type="sub_section">
      <SectionTitle>
2.1 Word Sense Dis~mhlguation
</SectionTitle>
      <Paragraph position="0"> Word sense disambiguation is an active research area in natural language processing, with a great number of novel methods proposed. Methods can typically be delineated along two dimensions, corpns-based vs. dictionary-based approaches.</Paragraph>
      <Paragraph position="1"> Corpus-based word sense disambignation algorjthm~ such as (Ng and Lee, 1996; Bruce and Wiebe, 1994; Yarowsky, 1994) relied on supervised learning fzom annotated corpora. The main drawback of these approaches is their requirement of a sizable sense-tagged corpus. Attempts to alleviate this tagbottleneck i~lude tmotstr~ias (Te~ ot ill,, 1996; Hearst, 1991) and unsupervised algorith~ (Yarowsky, 199s) Dictionary-based approaches rely on linguistic knowledge sources such as ma~l~i,~e-readable dictionaries (Luk, 1995; Veronis and Ide, 1990) and Word-Net (Agirre and Rigau, 1996; Resnik, 1995) and e0(ploit these for word sense disaznbiguation.</Paragraph>
      <Paragraph position="2"> Thus far, two notable sense-tagged corpora, the semantic concordance of WordN'et 1.5 (Miller et al.</Paragraph>
      <Paragraph position="3"> ,1994) and the DSO corpus of 192,800 sense-tagged occtuTences of 191 words used by (Ng and Lee, 1996) are still insu~cient in scale for supervised algorithms to perform well on a wide range of texts.</Paragraph>
      <Paragraph position="4"> Unsupervised algorit~m~ such as (Yarowsky, 1995) have reported good accuracy that rivals that of supervised algorithms. However, the algorithm was only tested on coarse-level senses and not on the refined sense distinctioas of WordNet, which is the required sense granularity of our approach.</Paragraph>
      <Paragraph position="5"> We hence turn to dictionary-based approaches, focusing on WordNet-based algorithms Since they fit in snugly with our WordNet-based semantic class disambiguation task.</Paragraph>
    </Section>
    <Section position="2" start_page="57" end_page="57" type="sub_section">
      <SectionTitle>
Information Content
</SectionTitle>
      <Paragraph position="0"> Resnik (1995) proposed a word sense disambiguation algorithm which determ~ the senses of words in noun groupings. The sense of a word is disambiguated by choosing the sense which is most highly supported by the other nouus of the noun group.</Paragraph>
      <Paragraph position="1"> The extent of support depends on the information content of the subsumers of the nouns in Word.Net, whereby information content is defined as negative log 1;1~1.~hood -togp(c), and p(c) is the probability of encountering an instance of concept c.</Paragraph>
      <Paragraph position="2"> As mentioned in his paper, although his approach was only reported on the disambiguation of words in related noun groupings, it can potentially be applied to word sense disambiguation of nouns in r-~-;~g text.</Paragraph>
      <Paragraph position="3"> In our implementation of his approach, we applied the method to general word sense disambiguation.</Paragraph>
      <Paragraph position="4"> We used the surrounding nouns of a word in free vmn~g text as the &amp;quot;norm grouping&amp;quot; and followed his algorit~r~ without modifications ~.</Paragraph>
    </Section>
    <Section position="3" start_page="57" end_page="57" type="sub_section">
      <SectionTitle>
Conceptual Density
</SectionTitle>
      <Paragraph position="0"> Agirre and Rigau:s (1996) approach has a ~imilar motivation as Kesnik's. Both approaches hinge on the belief that surrounding noun.~ provide strong clues as to the sense of a word.</Paragraph>
      <Paragraph position="1"> The main difference lies in how they determine the extent of support offered by the surrounding nouns.</Paragraph>
      <Paragraph position="2"> Agirre and Rigau uses the conceptual density of the ancestors of the nouns in WordNet as their metric.</Paragraph>
      <Paragraph position="3"> Our implementation foliow$ the pseudo-code pre-ZThe pseudo-code of his algorithm is detailed in (Res~ik, x995). =Surrounding nouns in the o~na\] ResnJk's approach refers to the other nouns in the noun grouping.</Paragraph>
      <Paragraph position="4"> umted in (Agirre and Rigan, 1996) s. For words which the algorithm failed to disambiguate (when no senses or more than one sense is returned), we relied on the most frequent heuristic.</Paragraph>
    </Section>
    <Section position="4" start_page="57" end_page="57" type="sub_section">
      <SectionTitle>
2.2 Semantic Distance
</SectionTitle>
      <Paragraph position="0"> The task of the semantic distance module is to reflect accurately the notion of &amp;quot;closeness&amp;quot; between the chosen concept node of the word and the semantic class nodes. It thus requires a metric which can effectively represent the semantic distance bet~veen two nodes in a taxonomy such as Word.Net.</Paragraph>
      <Paragraph position="1"> Conceptual Distance Rada et. al (1989) proposed such a metric termed as conceptual distance. Conceptual distance between two nodes is defined as the m~.ir-mn number of edges separating the two nodes. Take the example in Figure 2(b), the conceptual distance between &amp;quot;plane:l&amp;quot; and &amp;quot;aircraft:I&amp;quot; is 1, that between =plane:l&amp;quot; and &amp;quot;vehicle:l&amp;quot; is 2, and that between =plane:l&amp;quot; and &amp;quot;car:l&amp;quot; is 44.</Paragraph>
    </Section>
    <Section position="5" start_page="57" end_page="57" type="sub_section">
      <SectionTitle>
Link Probability
</SectionTitle>
      <Paragraph position="0"> The 11~1~ probability metric is our variant of the conceptual distance metric. Instead of considering all edges as equi-distance, the probability of the 1.1n\]C/ (or edge) is used to bias its distance. This metric is motivated by Resnik's use of the probability of instance occurrences of concepts, p(c) (Resnik, 1995).</Paragraph>
      <Paragraph position="1"> Link probability is defined as the difference between the probability of instance occurrences of the parent and child of the \]i.k~ Formally, Lin&amp;l'P(e, b) = p(a) - p(b), SWe clarified with the authoz~ certain parts of the algorithm which we find unclear. These axe the poin~ worth noting : null (1) corrtpu%e.concephtaLdens/b 9 of Step 2 only computes the conceptual density of concepts which are not ~-rked inva~d; (2) ex/%Ioop of Step 3 occurs whsu all senses subsumed by conce~ were already pzeviously disambiguatecl or when one or more senses of the word to be disambiguated are subsumed by con~elm~ (3) ~z~rLd~=r'n.5~r~zte&amp;ser~ of Step 4 marks senses subsumed by concept as disambiguated, marks concept and its clfddren as invalid, and discards other senses of the wor~ wi~ sere(s) disambiguated by C/on~; (4) disambiguated se~es of 'words which form the con- null text are not brought forward to the next window. 41.n Word.Net, these are 25 unique beginners of the taxonomy, instead of a co~on root. Hence, in our hnplementation, we ~.ign a large conceptual distance of 999 to the virtual edges between two unique beginners.</Paragraph>
      <Paragraph position="3"> ~h,e.re wm,ds((c)) i~ 'the C/eC/ o,f ~,0~.~ ~ the ~orlm,~ w~h ~re a~ba~med b~ the C/~..e~ (c), ~d 2V/# the ~o~Z ~mbee of ~to~ffi~ oC/~rr~/n ~e C/orp~, (~.e~/k, 1998) a m ~r~t~ of the linJ~, b m C/dt//.d o.f the link.</Paragraph>
      <Paragraph position="4">  The intuition behind this mewic is that the distance between the parent and the child should be &amp;quot;closer if the probability of the parent is close to that of the child, since that implies that whenever an instance of the parent occurs in the corpus, it is usually an instance of the child.</Paragraph>
    </Section>
    <Section position="6" start_page="57" end_page="57" type="sub_section">
      <SectionTitle>
Descendant Coverage
</SectionTitle>
      <Paragraph position="0"> In the same spirit, the descendant coverage mettic attempts to tweak the constant edge distance assumption of the conceptual distance metric. Instead of relying on corpus statistics, static inforn~.tion from Word.Net is exploited. Descendant coverage of a l~nlc is defined as the difference in the percentage of descendants subsumed by the parent and that subsumed by the child : null The same intuition underlies this metric; that the distance between the parent and the child should be &amp;quot;nearer&amp;quot; if the percentage of descendants subsumed by the parent is close to that of the child, since it indic~es that most descendants of the pare~ are also descendants of the child.</Paragraph>
      <Paragraph position="1"> Taxonomic Link (IS-A) All the metrics detailed above were designed to capture semantic similarity or closeness. The semantic class disambi~ion problem, however, is essentially to identify membership of the chosen concept node in the semantic class nodes.</Paragraph>
      <Paragraph position="2"> A simple implementation of the s~n~n~c distance module can thus be just a waversal of the taxonomic l~b~ (IS-A) of Word.Net. If the chosen concept node is s descendant of a s~n~=~ic class node, it should be classified as that s~a~tic class.</Paragraph>
    </Section>
  </Section>
class="xml-element"></Paper>