<?xml version="1.0" standalone="yes"?>
<Paper uid="C00-2148">
  <Title>Integrating compositional semantics into a verb lexicon</Title>
  <Section position="3" start_page="0" end_page="0" type="metho">
    <SectionTitle>
2 Levin classes
</SectionTitle>
    <Paragraph position="0"> Levin verb classes are based on the ability of a verb to occur in diathesis alternations, which are pairs of syntactic frames that are in some sense meaning preserving. The fundalnental assulnption is that the syntactic frames arc a direct reflection of the underlying semantics, ltowever, Levin classes exhibit inconsistencies that have hampered researchers' ability to reference them directly in applications. Many verbs ale listed in multiple classes, some of which have confl icting sets of syntactic frames. Dang et al.</Paragraph>
    <Paragraph position="1"> (1998) showed that multiple listings could in some cases be interpreted as regular sense extensions, and defined intersective Levin classes, which are a more line-grained, syntactically and semantically coherent refinement of basic Levin classes. We represent these verb classes and their regular sense extensions in the LTAG forlnalism.</Paragraph>
  </Section>
  <Section position="4" start_page="0" end_page="1011" type="metho">
    <SectionTitle>
3 Lexicalized rI?ee Adjoining Grammars
</SectionTitle>
    <Paragraph position="0"/>
    <Section position="1" start_page="0" end_page="1011" type="sub_section">
      <SectionTitle>
3.1 Overview of formalism
</SectionTitle>
      <Paragraph position="0"> Lexicatized Tree Adjoining Granunars consist of a finite set of initial and auxiliary elementary hees, and two operations to combine them. The minimal, non-recursive linguistic structures of a language, such as a verb and its complements, are captured by initial trees. Recursive structures of a language, such as prepositional modifiers which result in syntactically embedded VPs, are represented by auxiliary trees.</Paragraph>
      <Paragraph position="1">  Elementary trees are combined by the operations of substitution and adjunction. Substitution is a simple operation that replaces a leaf of a tree with a new tree. Adjunction is a splicing operation that replaces an internal node of an elementary tree with an auxiliary tree. Eveu tree is associated with a lexical item of the language, called the anchor of the tree.</Paragraph>
      <Paragraph position="2"> Tile tree represents the domain over which the lexical item can directly specify syntactic constraints, such as subject-verb number agreement, or semantic constraints, such as selectional restrictions, all of which are implemented as features.</Paragraph>
      <Paragraph position="3"> LTAGs are more powerful than context free grammars (CFG), allowing localization of so-called unbounded dependencies that cannot be handled by CFGs. There are critical benefits to lexical semantics that are provided by the extended domain of locality of the lexicalized trees. Each lexical entry corresponds to a tree. If the lexical item is a verb, the conesponding tree is a skeleton for an entire sentence with the verb already present, anchoring the tree as a terminal symbol. The other parts of the sentence will be substituted or adjoined in at appropriate places in the skeleton tree in the course of the derivation. The composition of trees during parsing is recorded in a derivation tree. The derivation tree nodes correspond to lexically anchored elementary trees, and the arcs are labeled with information about how these trees were combined to produce the parse. Since each lexically anchored initial tree corresponds to a semantic unit, the derivation tree closely resembles a semantic-dependency representation. null</Paragraph>
    </Section>
    <Section position="2" start_page="1011" end_page="1011" type="sub_section">
      <SectionTitle>
3.2 Semantics for TAGs
</SectionTitle>
      <Paragraph position="0"> There is a range of previous work in incorporating semantics into TAG trees. Stone and Doran (1997) describe a system used for generation that simultaneously constructs the semantics and syntax of a sentence using LTAGs. Joshi and Vijay-Shanker (1999), and Kallmeyer and Joshi (1999), describe the semantics of a derivation tree as a set of attachments of trees. The semantics of these attachments is given as a conjunction of formulae in a flat semantic representation. They provide a specific methodology for composing semantic representations much like Candito and Kahane (1998), where the directionality of dominance in the derivation tree should be interpreted according to the operations used to build it. Kallmeyer and Joshi also use a flat semantic representation to handle scope phenomena involving quantifiers.</Paragraph>
      <Paragraph position="1"> 4 Description of the verb lexicon VerbNet can be viewed in both a static and a dynamic way. Tile static aspect refers to the verb and class entries and how they are organized, providing the characteristic descriptions of a verb sense or a verb class (Kipper et al., 2000). The dynamic aspect of the lexicon constrains the entries to allow a compositional interpretation in LTAG derivation trees, representing extended verb meanings by incorporating adjuncts.</Paragraph>
      <Paragraph position="2"> Verb classes allow us to capture generalizations about verb behavioL Each verb class lists the tllemafic roles that the predicate-argument structure of its members allows, and provides descriptions of the syntactic fi'ames conesponding to licensed constructions, with selectional restrictions defined for each argument in each frame, l Each frame also includes semantic predicates describing the participants at various stages of the event described by the frame.</Paragraph>
      <Paragraph position="3"> Verb classes are hierarchically organized, ensuring that each class is coherent - that is, all its members have common semantic elements and share a common set of thematic roles and basic syntactic frames. This requires some manual restructuring of the original Levin classes, which is facilitated by using intersective Levin classes.</Paragraph>
    </Section>
  </Section>
  <Section position="5" start_page="1011" end_page="1013" type="metho">
    <SectionTitle>
5 Compositional Semantics
</SectionTitle>
    <Paragraph position="0"> We use TAG elementary trees for the description of allowable frames and associate semantic predicates with each tree, as was done by Stone and Doran. The semantic predicates are primitive enough so that many may be reused in different trees. By using TAGs we get the additional benefit of an existing parser that yields derivations and derived trees fiom which we can construct the compositional semantics of a given sentence.</Paragraph>
    <Paragraph position="1"> We decompose each event E into a tripartite structure in a manner similar to Moens and Steedman (1988), introducing a time function for each predicate to specify whether the predicate is true in the preparatory (d~ring(E)), cuhnination (er~d(E)), or consequent (res~ll:(E)) stage of an event.</Paragraph>
    <Paragraph position="2"> hfitial trees capture tile semantics of the basic senses of verbs in each class. For example, many IThese restrictions are more like preferences that generate a preferred reading of a sentence. They may be relaxed depending on the domain of a particular application.</Paragraph>
    <Paragraph position="4"> Figure 1 : Induced action alternation for the Run verbs verbs in the Run class can occur in the induced action alternation, in which the subject of the inmmsitive sentence has the same thematic role as the direct object in the transitive sentence. Figure l shows the initial trees for the transitive and intransitive variants for the Run class, along with their semantic predicates. The entity in motion is given by argl, associated with the syntactic subject of the intransitive tree and the direct object of the transitive tree. The event denoted by the transitive variant is a composition of two subevents: E1 refers to the event of av.ql running, and E2 refers to the event of an entity (argO) causing event El.</Paragraph>
    <Paragraph position="5"> Predicates are associated with not only the verb trees, but also the auxiliary trees. We use a flat semantic rcpmsentatiou like that of Kalhncycr and Joshi, and the semantics of a sentence is the conjunction of the semantic predicates of the trees used to derive the sentence. Figure 2 shows au auxiliary tree for a path prepositional pllrase headed by &amp;quot;to&amp;quot;, along with its associated semantic predicate.</Paragraph>
    <Paragraph position="6"> When the PP tree for &amp;quot;to the park&amp;quot; is adjoiued into the intransitive tree for &amp;quot;John ran&amp;quot;, the semantic interpretation is the conjunction of the two predicates motion(during(E),john) A goal(end(E),john, park); adjunction into the transitive tree for &amp;quot;Bill ran the horse&amp;quot; yields cause(during(E2),bilI, El) A motion(during(El), horse) A goal(end(El), horse, park).</Paragraph>
    <Paragraph position="7"> In both cases, the argument X,rsC/0.,rgl (john or horse) for the anxiliary tree is noulocal and colnes from the adjunction site. 2 The arguments are recovered from the derivation tree, following Candito and Kahane. When an initial tree is substituted into another tree, the dependency mirrors the derivation structure, so the variables associated with the sub2X.,..qo,.,.ga is the variable associated with the cntity in motion (ar91) in the tree to which tile PP a(Uoins (argO).</Paragraph>
    <Paragraph position="8"> stituting tree can be referenced as arguments in the host tree's predicates. When an auxiliary tree is adjoined, the dependency for the adjunction is reversed, so that variables associated with the host tree can be referenced as arguments in the adjoining tree's predicates.</Paragraph>
    <Paragraph position="10"> The tripartite event structure allows us to express the semantics of classes of verbs like change of state verbs whose description requires reference to a complex event structure. In the case of a verb such as &amp;quot;break&amp;quot;, it is important to make a distinction between the state of the object before the end of the action and the new state that results afterwards. This event structure also handles the eonative construction, in which there is an intention of a goal during the event, that is not achieved at the end of the event. The example of the conarive construction shown in Figure 3 expresses the intention of hitting something. Because the intention is not satisfied the semantics do not include the predicates manner(end(E),fi~rcefuI, X, rgo ) A conmct(end(E),X, rgo,Xc~rgO, that express the completion of the contact with impact event.</Paragraph>
    <Paragraph position="11"> The ability of verbs to take on extended senses in sentences based on their adjuncts is captured in a</Paragraph>
    <Paragraph position="13"> natural way by the TAG operation of adjunction and our conjunction of semantic predicates. The original Hit verb class does not include movement of the direct object as part of the meaning of hit; only sudden contact has to be established. By adjoining a path PP such as &amp;quot;across NP&amp;quot;, we get an extended meaning, and a change in Levin class membership to the Throw class. Figure 4 shows the class-specific auxiliary tree anchored by the preposition &amp;quot;across&amp;quot; together with its semantic predicates, introducing a motion event that immediately follows (meets) the contact event.</Paragraph>
    <Paragraph position="15"> oll the LTAG formalism, for which we already have a large English grammar. Palmer et al. (1998) defined compositional semantics for classes of verbs implemented in LTAG, representing general semantic components (e.g., motion, manner) as features on the nodes of the trees. Ore&amp;quot; use of separate logical forms gives a more detailed semantics for the sentence, so that for an event involving motion, it is possible to know not only that the event has a motion semantic component, but also which entity is actually in motion. This level of detail is necessary for applications such as animatiou of natural language instructions (Bindiganavale et al., 2000). Another important contribution of this work is that by dividing each event into a tripartite structure, we permit a more precise definition of the associated semantics.</Paragraph>
    <Paragraph position="16"> Finally, the operation of adjunction in TAGs provides a principled approach to representing the type of regular polysemy that has been a major obstacle in buildiug verb lexicons.</Paragraph>
    <Paragraph position="17"> Researching whether a TAG grammar for VerbNet can be automatically constructed by using development tools such as Xia et al. (1999) or Candito (1996) is part of our next step. We also expect to be able to factor out some class-specific auxiliary trees to be used across several verb classes.</Paragraph>
  </Section>
class="xml-element"></Paper>