<?xml version="1.0" standalone="yes"?>
<Paper uid="J90-4001">
  <Title>ANAPHORA RESOLUTION IN SLOT GRAMMAR</Title>
  <Section position="3" start_page="0" end_page="0" type="intro">
    <SectionTitle>
2 SLOT GRAMMAR
</SectionTitle>
    <Paragraph position="0"> The original work on Slot Grammar was done around 1976-1978 and appeared in McCord (1980). Recently, a new version (McCord 1989b, 1990) was developed in a logic programming framework, in connection with the machine translation system LMT (McCord 1989a, 1989c, 1989d).</Paragraph>
    <Paragraph position="1"> Slot Grammar is lexicalist and is dependency oriented. Every phrase has a head word (with a given word sense and morphosyntactic features). The constituents of a phrase besides the head word, also called the modifiers of the head, are obtained by &amp;quot;filling&amp;quot; slots associated with the head. Slots are symbols like subi, obi, and iobi representing grammatical relations, and are associated with a word Computational Linguistics Volume 16, Number 4, December 1990 197 Shalom Lappin and Michael McCord Anaphora Resolution in Slot Grammar (sense) in two ways. The lexical entry for the word specifies a set of complement slots, corresponding to logical arguments of the word sense, and the grammar specifies a set of adjunct slots for each part of speech. 4 A complement slot can be filled at most once, and an adjunct slot can by default be filled any number of times.</Paragraph>
    <Paragraph position="2"> The phenomena treated by augmented phrase structure rules in some grammatical systems are treated modularly by several different types of rules in Slot Grammar. The most important type of rule is the (slot) filler rule, which gives conditions (expressed largely through unification) on the filler phrase and its relations to the higher phrase.</Paragraph>
    <Paragraph position="3"> Filler rules are stated (normally) without reference to conditions on order among constituents. But there are separately stated ordering rules.5 Slot~head ordering rules state conditions on the position (left or right) of the slot (filler) relative to the head word. Slot~slot ordering rules place conditions on the relative left-to-right order of (the fillers of) two slots.</Paragraph>
    <Paragraph position="4"> A slot is obligatory (not optional) if it must be filled, either in the current phrase or in a raised position through left movement or coordination. Adjunct slots are always optional. Complement slots are optional by default, but they may be specified to be obligatory in a particular lexical entry, or they may be so specified in the grammar by obligatory slot rules. Such rules may be unconditional or be conditional on the characteristics of the higher phrase.</Paragraph>
    <Paragraph position="5"> They also may specify that a slot is obligatory relative to the filling of another slot. For example, the direct object slot in English may be declared obligatory on the condition that the indirect object slot is filled by a noun phrase.</Paragraph>
    <Paragraph position="6"> One aim of Slot Grammar is to develop a powerful language-independent module, a &amp;quot;shell,&amp;quot; which can be used together with language-dependent modules, reducing the effort of writing grammars for new languages. The Slot Grammar shell module includes the parser, which is a bottom-up chart parser. It also includes most of the treatment of coordination, unbounded dependencies, controlled subjects, and punctuation. And the shell contains a system for evaluating parses, extending Heidorn's (1982) parse metric. The Slot Grammar evaluator is used not only for ranking final parses, as with Heidorn's, but also for pruning away unlikely partial analyses during parsing, thus reducing the problem of parse space explosion. Parse evaluation expresses preferences for close attachment, for choice of complements over adjuncts, and for parallelism in coordination. null Although the shell contains most of the treatment of the above phenomena (coordination, etc.), a small part of their treatment is necessarily language dependent. A (languagespecific) grammar can include for instance (1) rules for coordinating feature structures that override the defaults in the shell; (2) declarations of slots (called extraposer slots) that allow left extraposition of other slots out of their fillers; (3) language-specific rules for punctuation that override defaults; and (4) language-specific controls over parse evaluation that override defaults.</Paragraph>
    <Paragraph position="7"> Currently, Slot Grammars are being developed for English, (ESG) by McCord, for Danish (DSG) by Arendse Bernth, and for German (GSG) by Ulrike Schwall. ESG uses two lexicons: (1) a hand-coded lexicon of about 3,700 common words, and (2) the UDICT lexicon (Byrd 1983; Klavans and Wacholder 1989) having over 60,000 lemmas, with a heuristic interface that produces Slot Grammarstyle entries.</Paragraph>
    <Paragraph position="8"> Our anaphora algorithms apply in a second pass to the parse output; the remainder of this section describes Slot Grammar syntactic analysis structures.</Paragraph>
    <Paragraph position="9"> A syntactic structure is a tree; each node of the tree represents a phrase in the sentence and has a unique head word. Formally, a phrase is represented by a term phrase(X,H, Senso,Features,glotFramo,E~,Mods), where the components are as follows. (I) X is a logical variable called the marker of the phrase. Unifications of the marker play a crucial role in the anaphora algorithms.</Paragraph>
    <Paragraph position="10"> (2) H is an integer representing the position of the head word of the phrase. This integer identifies the phrase uniquely, and is used in the anaphora algorithms as the way of referring to phrases. (3) Sense is the word sense of the head word. (4) Features is the feature structure of the head word and of the phrase. It is a logic term (not an attribute-value list), which is generally rather sparse in information, showing mainly the part of speech and inflectional features of the head word. (5) SlotFrame is the list of complement slots, each slot being in the internal form slot(Slot,Ob,X), where Slot is the slot name, Ob shows whether it is an obligatory form of Slot, and X is the slot marker. The slot marker is unified (essentially) with the marker of the filler phrase when the slot is filled, even remotely, as in left movement or coordination. Such unifications are important for the anaphora algorithms. (6) Ext is the list of slots that have been extraposed or raised to the level of the current phrase. (7) The last component Mods represents the modifiers (daughters) of the phrase, and is of the form mods(LMods,RMods) where LMods and RMods are the lists of left modifiers and right modifiers, respectively. Each member of a modifier list is of the form Slot:Phrase where Slot is a slot and Phrase is a phrase that fills Slot. Modifier lists reflect surface order, and a given slot may appear more than once (if it is an adjunct). Thus modifier lists are not attribute-value lists.</Paragraph>
    <Paragraph position="11"> Figure I shows a sample parse produced by ESG for the sentence Who did John say wanted to try to find him? The tree is displayed by a procedure that uses only one line per node and exhibits tree structure lines on the left. In this display, each line (representing a node) shows (1) the tree connection lines, (2) the slot filled by the node, (3) the word sense predication, and (4) the feature structure. The feature structure is abbreviated here by a display option, showing only the part of speech. The word sense predication consists of the sense name of the head word with the 198 Comimtational Linguistics Volume 16, Number 4, December 1990 Shalom Lappin and Michael McCord Anaphora Resolution in Slot Grammar Who did John say wanted to try to find him?</Paragraph>
    <Paragraph position="13"> following arguments. The first argument is the marker variable for the phrase (node) itself; it is like an event or state variable for verbs. The remaining arguments are the marker variables of the slots in the complement slot frame (u signifies &amp;quot;unbound&amp;quot;). As can be seen in the display, the complement arguments are unified with the marker variables of the filler complement phrases. Note that in the example the marker X2 of the 'who' phrase is unified with the subject variables of &amp;quot;want,&amp;quot; &amp;quot;try,&amp;quot; and &amp;quot;find.&amp;quot; (There are also some unifications created by adjunct slot filling, which will not be described here.) For the operation of our anaphora algorithms, there is a preliminary step in which pertinent information about the parse tree is represented in a more convenient way for the algorithms. As indicated above, nodes (phrases) themselves are represented by the word numbers of their head words.</Paragraph>
    <Paragraph position="14"> Properties of phrases and relations between them are represented by unit clauses (predications) involving these integers (and other data), which are asserted into the Prolog workspace. Because of this &amp;quot;dispersed&amp;quot; representation with a collection of unit clauses, the original phrase structure for the whole tree is first grounded (variables are bound to unique constants) before the unit clauses are created.</Paragraph>
    <Paragraph position="15"> As an example of this clausal representation, the clause hasarg(P,X) says that phrase P has X as one of its arguments; i.e., X is the slot marker variable for one of the complement slots of P. For the above sample parse, then, we would get clauses hasarg(S,'X2'), hasarg(5,'X12').</Paragraph>
    <Paragraph position="16"> as information about the 'want' node (5).</Paragraph>
    <Paragraph position="17"> As another example, the clause phmarker(P,X) is added when phrase P has marker X. Thus for the above sample, we would get the unit clause phmarker(1,'X2').</Paragraph>
    <Paragraph position="18"> An important predicate for our algorithms is pharg, defined by pharg(P, Q) ~ phmarker(P,X) &amp; hasarg(Q,X).</Paragraph>
    <Paragraph position="19"> This says that phrase P is an argument of phrase Q. This includes remote arguments and controlled subjects, because of the unifications of marker variables performed by the Slot Grammar parser. Thus for the above parse, we would get pharg(l,5), pharg(1,7), pharq(1,9).</Paragraph>
    <Paragraph position="20"> showing that &amp;quot;who&amp;quot; is an argument of &amp;quot;want,&amp;quot; &amp;quot;try,&amp;quot; and &amp;quot;find.&amp;quot;</Paragraph>
  </Section>
class="xml-element"></Paper>