File Information
File: 05-lr/acl_arc_1_sum/cleansed_text/xml_by_section/metho/89/e89-1000_metho.xml
Size: 18,731 bytes
Last Modified: 2025-10-06 14:12:14
<?xml version="1.0" standalone="yes"?> <Paper uid="E89-1000"> <Title>Current Issues in Computational Lexical Semantics</Title> <Section position="4" start_page="264" end_page="264" type="metho"> <SectionTitle> 2. Methods in Lexical Semantics </SectionTitle> <Paragraph position="0"> Given what I have said, let us examine the questions presented above in more detail. Because of space, I will have little to say concerning question (4) in this short note. First, let us turn to the issue of methodology. How can we determine the soundness of our method? Are new techniques available now that have not been adequately explored? Very briefly, we can summarize the most essential techniques assumed by the field, in some way, as follows (see, for example \[Cruse, 1986\]): 1. On the basis of categorial distinctions, establish the fundamental differences between the grammatical classes; the typicial semantic behavior of a word of category X.</Paragraph> <Paragraph position="1"> 2. Find distinctions between elements of a word class on the basis of collocation and cooccurrence tests. For example, &quot;dog&quot; and &quot;rock&quot; partition into different classes due to certain features. 3. Test for distinctions of a grammatical nature on the basis of diathesis; i.e. alternations that are realized in the syntax. For example, break vs. cut in (i) and (ii) below \[Hale and Keyser, 1986\]. (i) a. The glass broke.</Paragraph> <Paragraph position="2"> b. John broke the glass.</Paragraph> <Paragraph position="3"> (ii) a. *The bread cut.</Paragraph> <Paragraph position="4"> b. John cut the bread.</Paragraph> <Paragraph position="5"> - xviii This results in ever-finer distinctions in the semantic and syntactic behavior of verbs. 4. Test for entailments in the word senses of a word, in different grammatical contexts. The same lexical item may carry different entailments in different contexts. For example, &quot;forget&quot; in (i): (i) a. John forgot that he locked the door.</Paragraph> <Paragraph position="6"> b. John forgot to lock the door.</Paragraph> <Paragraph position="7"> Sentence (a) has a factive interpretation of &quot;forget&quot; that (b) does not carry. 5. Test for the ambiguity of a word. Distinguish between ambiguity and polysemy, (cf. \[Hirst, 1987\], \[Fass, 1988\]) and from the accidental nature and the logical nature of ambiguity. For example, the ambiguity between the two senses of &quot;bank&quot; in (i) and the polysemy in (ii).</Paragraph> <Paragraph position="8"> (i) a. the bank of the river b. the richest bank in the city (ii) a. The bank raised its interest rates yesterday.</Paragraph> <Paragraph position="9"> b. The store in next to the bank.</Paragraph> <Paragraph position="10"> 6. Establish what the compositional nature of a lexical item is when applied to other words. For example, &quot;alleged&quot; vs. &quot;tall&quot; in the example below.</Paragraph> <Paragraph position="11"> (i) a. the alleged suspect b. the tall suspect While &quot;tall&quot; is a simple intersective modifier in (b) we must be aware of the intensional behavior of adjectives such as &quot;alleged&quot; in (a).</Paragraph> <Paragraph position="12"> This just briefly characterizes some of the techniques that have been useful for arriving at pre-theoretic notions of word meaning. What has changed over the years are not so much the methods themselves as the descriptive details provided by each test. One thing that has changed, however, --and this is significant-- is the way computational lexicography has provided stronger techniques and even new tools for lexical semantics research: for sense discrimination tasks \[Atldns 1987\]; for constructing concept taxonomies \[Amsler 1985, Atldns, Klavans and Boguraev, forthcoming\]; for establishing semantic relatedness among word senses \[Wilks et al, 1988\]; as well as for testing new ideas about semantic representations \[Boguraev and Pustejovsky, forthcoming\].</Paragraph> </Section> <Section position="5" start_page="264" end_page="264" type="metho"> <SectionTitle> 3. Descriptive Adequacy of Existing Representations </SectionTitle> <Paragraph position="0"> Turning now to the question of how current theories compare with the coverage of lexical semantic data, there are two generalizations that should be made. First, the taxonomic descriptions that have been made of verb classes are far superior to the classifications available twenty years ago (see \[Levin, 1985\] for review). Using mainly the descriptive vocabulary of \[Talmy, 1975, 1985\] and \[Jackendoff, 1983\], fine and subtle distinctions are drawn that were impossible to capture in the earlier primitives-approach of \[Schank, 1972, 1975\] or the frame semantics of \[Fillmore, 1968\]. There are two problems with the current work, however. Unlike the theories of \[Wilks, 1975\] and \[Quillian, 1968\], there is no general coherent view on what the entire lexicon will look like when semantic structures for other major categories are studied. Furthermore, it is not clear whether this work embodies a real theory of knowledge representation. We can, however, view this research as the first step towards constructing a theory of lexical semantics which is integrated into a linguistic theory, as well as interpreted in a knowledge representation system.</Paragraph> </Section> <Section position="6" start_page="264" end_page="264" type="metho"> <SectionTitle> - xix - 4. Explanatory Adequacy of Existing Representations </SectionTitle> <Paragraph position="0"> Finally, let us turn to the question of whether current theories have changed the way we look at representation and lexicon design. The question here is whether the representations assumed by current theories are adequate to account for the richness of natural language semantics. It should be pointed out here that a theory of lexical meaning will affect the general design of our semantic theory in several ways. If we view the goal of a semantic theory as being able to recursively assign meanings to expressions, (accounting for phenomena such as synonymy, antonymy, polysemy, metonymy, etc.), then our view of compositionality depends ultimately on what the basic lexical categories of the language denote. Conventional wisdom on this point paints a picture of words behaving as either active functors or passive arguments \[Montague, 1974\]. But if we change the way which categories can denote, then the form of compositionality itself changes. Therefore, if done well, lexical semantics forces us to reevaluate the very nature of semantic composition in language (see, for example \[Keenan and Faltz, 1985\]).</Paragraph> <Paragraph position="1"> In what ways could lexical semantics affect the larger methods of composition in semantics? I mentioned above that most of the careful representation work has been done on verb classes. In fact, the semantic weight in both lexical and compositional terms usually falls on the verb. This has obvious consequences for how to treat lexical ambiguity. For example, consider the verb &quot;bake&quot; in the two sentences below.</Paragraph> <Paragraph position="2"> (1) a. John baked the potato.</Paragraph> <Paragraph position="3"> b. John baked the cake.</Paragraph> <Paragraph position="4"> \[Atkins, Kegl, and Levin, 1988\] demonstrate that verbs such as &quot;bake&quot; are ambiguous, with both a &quot;change-of-state&quot; sense (la) and a &quot;create&quot; sense (lb).</Paragraph> <Paragraph position="5"> A similar ambiguity exists with verbs that allow the resulative construction, shown in (2) and (3), and discussed in \[Dowty, 1979\], \[Jackendoff, 1983\], and \[Levin and Rapoport, 1988\].</Paragraph> <Paragraph position="6"> (2) a. Mary hammered the metal.</Paragraph> <Paragraph position="7"> b. Mary hammered the metal flat.</Paragraph> <Paragraph position="8"> (3) a. John wiped the table.</Paragraph> <Paragraph position="9"> b. John wiped the table clean.</Paragraph> <Paragraph position="10"> On many views, the verbs in (2) and (3) are ambiguous, related by either a lexical transformation \[Levin and Rapoport, 1988\], or a meaning postulate \[Dowty, 1979\]. In fact, given strict requirements on the way that a verb can project its lexical information, the verb &quot;run&quot; in (4) will also have two lexical entries, depending on the syntactic environment it selects \[Talmy, 1985\], \[Levin and Rappaport, 1989\]. (4) a. Mary ran yesterday.</Paragraph> <Paragraph position="11"> b. Mary ran to the store yesterday.</Paragraph> <Paragraph position="12"> These two verbs differ in their semantic representations, where the &quot;run&quot; in (4b) means &quot;go-to-bymeans-of-running&quot;, while in (4a) it means simply &quot;move-by-running&quot;. The methodology described above for distinguishing word senses is also assumed by those working in more formal frameworks. For example, \[Dowty, 1985\] proposes multiple entries for control and raising verbs, and establishes their semantic equivalence with the use of meaning postulates. That is, the verbs in (5) and (6) are lexically distinct but semantically related by rules.</Paragraph> <Paragraph position="13"> (5) a. It seems that John likes Mary.</Paragraph> <Paragraph position="14"> b. John seems to like Mary.</Paragraph> <Paragraph position="15"> - XX (6) a. Mary prefers that she come.</Paragraph> <Paragraph position="16"> b. Mary prefers to come.</Paragraph> <Paragraph position="17"> Given the conventional notions of function application and compositionality, there is little choice but to treat all of the above cases as polysemous verbs. Yet, something about the systematicity of such ambiguity suggests that a more general, simpler explanation should be possible. By relaxing the conditions on how the meaning of a complex expression is derived from its parts, we can, in fact, arrive at a very straightforward explanation for cases of &quot;logical polysemy.&quot; The basic idea is this. Rather than treating the expressions that behave as arguments to a function as simple, passive objects, imagine that they are as active in the semantics as the verb itself. The product of function application would be sensitive to both the function and its active argument. Something like this is suggested in \[Keenan and Faltz, 1985\], as the &quot;Meaning-Form Correlation Principle.&quot; What I have in mind can best be illustrated by returning to the examples in (1). Rather than having two word senses for a verb such as &quot;bake,&quot; suppose there is simply one, a &quot;change-of-state.&quot; Now, assume that &quot;bake&quot; can be lexically specified as a Process verb \[Pustejovsky, 1988\]. It is a general property of Processes, that they can shift their &quot;aspectual type&quot; to become a telic event. For example, &quot;push the wagon&quot; can shift from a Process to a Transition (&quot;push the wagon to Mary&quot;) by general principles of event composition (cf. \[Pustejovsky, 1988\]). Assume further, that our semantic theory makes a logical distinction between &quot;potatoes&quot; and &quot;cakes&quot; in terms of natural versus artifactual objects. Any artifact can be identified with the state of being that object, along with the purpose of the object, and other information. Most importantly, however, an artifact has a kind of stative interpretation as part of its meaning.</Paragraph> <Paragraph position="18"> Notice that the two senses for &quot;bake&quot; in (1) are the result of compositionality in a richer sense. For observe that the &quot;create&quot; sense arises by the same principle of event composition mentioned above. Namely, that there is a baking process, and it results in a cake. This is the &quot;create&quot; sense, and it is present just in those cases where an artifactual stative is already encoded in the argument (see \[Pustejovsky, 1989\] for details).</Paragraph> <Paragraph position="19"> In fact, similar principles seem to be operating in the examples mentioned in (2) through (4); namely, systematic ambiguity is the result of principles of semantic composition, rather than lexical ambiguity. I will call this particular kind of case &quot;cocompositionality.&quot; In explaining the behavior of the systematic ambiguity above, I made reference to properties of the Noun Phrase that are not typical semantic properties for nouns in linguistics, e.g. artifact, natural kind. In \[Pustejovsky, 1989\] and \[Pustejovsky and Anick, 1988\], I suggest that there is a system of relations that characterizes the semantics of nominals, very much like the argument structure of a verb. I call this the &quot;Qualia Structure&quot;, adopting a term from \[Goodman, 1951\]. Essentially, the qualia structure of a noun determines its meaning as much as the list of arguments determines a verb's meaning. The elements that make up a qualia structure include notions such as &quot;container,&quot; &quot;space,&quot; &quot;surface,&quot; &quot;figure,&quot; &quot;artifact,&quot; and so on.</Paragraph> <Paragraph position="20"> When we combine the qualia structure of a NP with the argument structure of a verb, we begin to see a richer notion of compositionality emerging, one that looks very much like object-oriented approaches to programming \[Ingria and Pustejovsky, 1988\]. As an example, consider the two senses of &quot;window&quot; and &quot;door&quot; in (7) and (8) below: (7) a. John crawled through the window.</Paragraph> <Paragraph position="21"> b. The window is closed.</Paragraph> <Paragraph position="22"> (8) a. Mary painted the door.</Paragraph> <Paragraph position="23"> b. Mary walked through the door.</Paragraph> <Paragraph position="24"> C/ - xxi - null Each noun appears to have two senses: an object denotation, and a portal denotation. We can characterize the meaning of such &quot;Double Figure Ground&quot; nominals as a three place relation, roughly something like N(Figure,Invert-Figure, Ground), where Figure refers to the artifact itself, Invert-Figure refers to the portal, and Ground is a reference constant. The foregrounding or backgrounding of a nominal's qualia is very similar to argument structure-changing operations for verbs. That is, in (Sa) &quot;paint&quot; applies to the Figure interpretation of &quot;the door,&quot; while in (8b), &quot;through&quot; will apply to the Invert-Figure interpretation of the same NP. The ambiguity with such nouns is a logical one, one that is intimately linked to the semantic representation of the object itself. The qualia structure is a way of capturing this fact.</Paragraph> <Paragraph position="25"> Having discussed the notion of logical polysemy, let us conclude our brief discussion of lexical semantics with the issue of metonymy. Metonymy, where a subpart or related part of an object &quot;stands&quot; for the object itself, also poses a problem for standard denotational theories of semantics. For example, how can we account for the reference shifts such as those shown in (9)? (9) a. Thatcher vetoed the channel tunnel.</Paragraph> <Paragraph position="26"> b. Mary enjoyed the book.</Paragraph> <Paragraph position="27"> I suggest that these are cases of semantic &quot;type coercion&quot;, where the verb has coerced the meaning of a term phrase into a different semantic type. In the case of (9a), it is obvious that what is vetoed is some proposal relating to the object. In (9b), the book is enjoyed only by virtue of some event or process that involves the book, performed by Mary. It might furthermore be reasonable to assume that the qualia structure of &quot;book&quot; specifies what the artifact is used for; i.e. reading. Such coercion results in a word sense for the NP that I will call &quot;logical metonymy.&quot; Roughly, logical metonymy is where a logical argument of a semantic type (selected by a function) denotes the semantic type itself. For details see \[Pustejovsky, 1989\].</Paragraph> <Paragraph position="28"> Another interesting example involves the possible subjects of causative verbs. These are also cases of logical metonymy. Consider the sentences in (10) and (11).</Paragraph> <Paragraph position="29"> (10) a. The pictures of himself frightened John.</Paragraph> <Paragraph position="30"> b. A cup of coffee in the morning woke John up.</Paragraph> <Paragraph position="31"> (11) a. John killed Mary.</Paragraph> <Paragraph position="32"> b. The gun killed Mary.</Paragraph> <Paragraph position="33"> c. John's stupidity killed Mary.</Paragraph> <Paragraph position="34"> d. The war killed Mary.</Paragraph> <Paragraph position="35"> e. John's pulling the trigger killed Mary.</Paragraph> <Paragraph position="36"> The sentences in (10) illusffate the various syntactic consequences of metonymy and coercion involving experiencer verbs, while those in (11) show the different &quot;metonymic extensions&quot; possible from the causing event in a killing. The generalization here is that when a verb selects an event as one of its arguments, type coercion to an event will permit a limited range of logical metonymy.</Paragraph> </Section> <Section position="7" start_page="264" end_page="264" type="metho"> <SectionTitle> 5. Conclusion </SectionTitle> <Paragraph position="0"> In this paper I have tried to cover what I believe are currently some of the most interesting issues in lexical semantics. I have also tried to outline an approach to meaning where the semantic load is spread more evenly throughout the lexicon. This entails a semantic description of noun classes, termed a &quot;theory of Qualia,&quot; as well as the rules of composition that allow this information to be incorporated into the semantic interpretation of larger expressions. Although much of what I have presented is incomplete and somewhat programmatic, I firmly believe this approach can help clarify the nature of word meaning and compositionality in natural language.</Paragraph> </Section> <Section position="8" start_page="264" end_page="264" type="metho"> <SectionTitle> - xxii - Acknowledgements </SectionTitle> <Paragraph position="0"> I would like to thank Dan Fass for lengthy discussion on many of the issues mentioned here. I would also like to thank Branimir Boguraev and Robert Ingria for useful comments. Any errors, of course, are my own.</Paragraph> </Section> class="xml-element"></Paper>