<?xml version="1.0" standalone="yes"?>
<Paper uid="T75-1004">
  <Title>YNTACTIC PROCESSING AND ~UNCTIONAL SENTENCE PERSPECTIVE</Title>
  <Section position="1" start_page="0" end_page="0" type="metho">
    <SectionTitle>
~YNTACTIC PROCESSING AND
~UNCTIONAL SENTENCE PERSPECTIVE
Martin Kay
XEROX PARC
</SectionTitle>
    <Paragraph position="0"> This paper contains some ideas that nave occurred to me in the course of some preliminary work on the notion of reversible grammar. In order to make it possible to generate and analyze sentences with the same grammar, represented in the same way, I was led to consider restrictions on the expressive power of the formalism that would be acceptible only if the structures of sentences contained more information than American linguisits have usually been prepared to admit. I hope to convey to you some of my surprise and delight in finding that certain linguists of the Prague school argue for the representation of this same information on altogether different grounds.</Paragraph>
    <Paragraph position="1"> I. REVERSIBLE GRAMMARS For me, a grammar is not so much a set of well-formedness conditions on strings of words as a relation between strings of words (sentences) and structures. Usually it is constructive in the sense that there is a device that interprets it to convert either structures into sentences (a generator) or sentences into structures (a parser). A grammar for which both a generator and a parser can be found is what I call reversible and, of course, what I am interested in is not so much particular reversible grammars as a formalism which guarantees reversibility for any grammar that adheres to it. Context-free grammars are clearly reversible in this sense and transformational grammars are clearly not.</Paragraph>
    <Paragraph position="2"> Augmented Transition Networks (ATNs) are reversible provided some quite reasonable restrictions are placed on the operations that can be performed on registers. This is not to say that it is a trivial matter to obtain a generator from an arbitrary ATN grammar.</Paragraph>
    <Paragraph position="3"> It goes without saying that the composition of a generator and a parser interpreting the same reversible grammar on the same sentence or structure does not, in general, perform the identity transformation. Frequently, sentences are ambiguous and structures often correspond to sets of more than one paraphrase. Parsing followed by generation of each member of the resulting set of structures will therefore yield the grammatical paraphrases of the original under all grammatical interpretations.</Paragraph>
    <Paragraph position="4"> The practical advantages of a reversible grammar are obvious. A computer program that engages in any kind of conversation must both parse and generate and it would be economical to base both processes on the same grammar. But, to me, the theoretical appeal is stronger. It is plausible that we have something in our heads that fills the function I am ascribing to grammar, though I am not insensitive to the claims of those who deny this. But it is altogether implausible that we have two such things, one for parsing and one for generation, essentially unrelated to one another.</Paragraph>
    <Paragraph position="5"> A left-to-right generator is one that deposits words into the output string one after another, in left-to-right sequence. A left-to-right parser is one that examines the words of the input string one after another, from left to right. A left-to-right reversible grammar is one for which there is a left-to-right generator and a left-to-right parser. Once again, it is clear that context-free grammars, in the usual notation, meet the requirements. ATN grammars probably do not. They certainly do not if we exclude the possibility of entirely reworking the grammar and presenting it in an entirely new form to the generator. The kind of grammar I have in mind would require no such reworking.</Paragraph>
    <Paragraph position="6"> Intuitively, the notion is a simple one. It is very like an ATN in that it analyses sentences by moving through a network, examining the input string at each transition and, if the current symbol meets the conditions specified for the transition, assigning it to a register. The generator would also move through the network, making exactly the same transitions, but depositing the contents of registers into the string at each step.</Paragraph>
  </Section>
  <Section position="2" start_page="0" end_page="0" type="metho">
    <SectionTitle>
2. THE PROCESSOR
</SectionTitle>
    <Paragraph position="0"> Generators and parsers for the kind of reversible grammar I have in mind could be implemented in a great variety of ways. One of the simplest I know would be to use a version of the General Syntactic Processor (GSP). GSP contains: (I) a grammar in the form of a transition network, that is, a airected graph in which the wermissible transitions between states are represented by arcs, each labeled with a more or less complicated set of actions which eetermine the applicability of the arc and cause side effects, such as the assignment of values to re~isters,  (2) an agenda of tasks to be carried out, (3) a chart, that is, a directed graph consisting of Y=ertiees and edges which represents the sentence being  analyzed or generated together with 1~s component parts--phrases, In~ermediate derivations, or whatever--which, together with the agenda, completely encapsulates the state of the entire processor at any glven point in its operation, (4) a set of schedulin~ rules whose job is to determine the order in which the tasks on the agenda will be carried out, ana (5) the interpreter +-~self.</Paragraph>
    <Paragraph position="1"> Edges in the chart are either complete or incomplete. Complete edges represent 12.</Paragraph>
    <Paragraph position="2"> completely specified words or phrases and, if there is a path through the chart from one edge to another, it is because the first precedes the second in temporal, or left-to-right, sequence. If there is no path from one to another, then they belong to alternative hypotheses about the structure of of the sentence. So, the sentence &amp;quot;they are flying planes&amp;quot; has, let us say, two analyses, each consisting of a noun phrase followed by a verb phrase. But there is no path between the phrases in one analysis and those in the other. The verb phrase in one analysis consists of the verb &amp;quot;are&amp;quot; followed by the noun phrase &amp;quot;flying planes&amp;quot;, which are therefore adjacent on the same path, but there is no path from either of them to the verb phrase they make up because this is an alternative analysis of the same set of words.</Paragraph>
    <Paragraph position="3"> An incomplete edge represents part of a phrase together with an indication of what would have to be added to complete it. For example, an incomplete noun phrase might include the string &amp;quot;the big black&amp;quot; plus an indication that a following noun, possibly preceded by some more adjectives, would complete it. A special kind of incomplete edge is an empty edge. Empty edges are successors of themselves. In other words, they are always incident from and to the same vertex, reflecting the fact that they represent no part of the sentence, but merely the potential for finding some structural component. The specification of how an incomplete edge can be completed takes the form of one or more arcs in the grammar, each paired with a direction--left or right. If the direction is right, then completion can be achieved by following sequences of arcs incident fro~ the given state; if it is left, then arcs incident to the state must be followed.</Paragraph>
    <Paragraph position="4"> The interpreter uses the scheduling rules to chose an item on the agenda which it then carries out. If all the tasks that were ever put on the agenda in the course of generation or analysis were carried out in an arbitrary order, then all results that the grammar allowed would be obtained sooner or later. The scheduling rules formalize strategie~ of one kind and another. They are presumably designed so as to shorten the time required to reach a result which is, in some sense, acceptable, at which time the remaining entries on the agenda can simply be abandoned.</Paragraph>
    <Paragraph position="5"> The typical task is an attempt to apply an arc from the grammar to an edge in the chart. If the arc applies successfully, ~ome new material will be added to the chart. In generation, the new material typically consists of one or two new edges constituting a sequence with the same end points as those of the initial edge.</Paragraph>
    <Paragraph position="6"> usually, no more than one of the newly introduced edges will be incomplete. Thus, there might be a task in which an arc was applied to the noun-phrase edge representing &amp;quot;the big black dog&amp;quot; and which resulted in the complete article &amp;quot;the&amp;quot; and the incomplete noun phrase &amp;quot;big black dog&amp;quot;. In parsing, the task specifies one or two i3.</Paragraph>
    <Paragraph position="7"> euges, one of which is incomplete. The idea is to attempt to take the complete edge at least one step nearer to completion by incorporating the other edge. If only one edge is specified, then the new edge will have the same end points as the original, but will presumably be differently labeled. Within this version of GSP, top-to-bottom, left-to-right parsing, in the manner of an ATN parser, proceeds broadly as follows: I. Whenever, as the result of introducing a new edge into the chart, a new sequence consisting of an incomplete eage followed by a complete one comes into existance, put a new task on the agenda for each of the &amp;quot;category&amp;quot; arcs named on the incomplete edge. When one of these tasks is executed, the arc will be applied to the complete e e giving rise, if successful, to a  new edge~ complete or incomplete.</Paragraph>
    <Paragraph position="8"> 2. Whenever a new incomplete edge is Introduced that names a &amp;quot;Pop&amp;quot; or &amp;quot;Jump&amp;quot; arc, create tasks that will cause these to be carried out.</Paragraph>
    <Paragraph position="9"> 3. Place an empty sentence edge before  the first word of the sentence.</Paragraph>
    <Paragraph position="10"> The process starts with step 3, which immediately causes a sequence of instances u~ steps I and 2.</Paragraph>
    <Paragraph position="11"> An incomplete edge represents a stack frame in the ATN processor. It is labeled with a set of registers and it spans a portion of the chart representing the part of the string so far analyzed. &amp;quot;Category&amp;quot; arcs are applied to an incomplete edge and a complete edge immediately to its right. If successful, the result is a new incomplete edge. &amp;quot;Pop&amp;quot; arcs produce a complete edge, in exchange for an incomplete one. &amp;quot;Jump&amp;quot; arcs produce an incomplete edge in exchange for an incomplete edge, the differences being in the arcs that specify how to proceed towards completion, and possibly in the label.</Paragraph>
    <Paragraph position="12"> It turns out that the mechanism of Fecursive calls that &amp;quot;Push&amp;quot; and &amp;quot;Pop&amp;quot; arcs provide for is embraced by the devices already described. Suppose that sentences are to be analyzed as consisting of a noun phrase followed by a verb phrase and that a noun phrase, say &amp;quot;the big black dog&amp;quot; has somehow been recognized at the beginning of the sentence. This means that there will be a complete edge representing this noun phrase and an incomplete sentence edge which has the same end points with an arc specifying that a verb phrase with a singular, third-person verb, is to follow. Since the grammar contains a subnetwork giving the structure of verb phrases, an empty edge labeled with the category &amp;quot;verb phrase&amp;quot; is introduced following that incomplete sentence provided there is not one already there. In due course, this will presumably cause a complete verb phrase to appear. The general principle is this: whenever an incomplete edge specifies a &amp;quot;category&amp;quot; arc for wnlch ~nere is a corresponding subnetwork, an empty edge is created following that one for each of the initial arcs in the subnetwork in the hope that this will lead to the creation of a new complete edge that the &amp;quot;category&amp;quot; arc can be successfully applied to.</Paragraph>
  </Section>
  <Section position="3" start_page="0" end_page="0" type="metho">
    <SectionTitle>
3. THE USE OF REGISTERS
</SectionTitle>
    <Paragraph position="0"> The principal problem with this simple plan, when applied to reversible grammars, is that the registers cannot be guaranteed to have the necessary contents at the time required. One of the strengths of the ATN formalism is that it allows the parser to &amp;quot;change its mind&amp;quot;. The canonical example is the passive construction. The first verb phrase in the sentence is assigned to the subject register. But when a passive verb--part of the verb &amp;quot;be&amp;quot; and a transitive past participle--has been encountered, the contents of the subject register are simply transferred to the object register. If a &amp;quot;by&amp;quot; phrase follows, its object will later go into the subject register. In this way, a great deal of backing up is avoided.</Paragraph>
    <Paragraph position="1"> In generating a passive sentence, it is clear that the first step cannot be to deposit the contents of the subject register in the first position. An alternative might be to decide which register to use in filling the first position by examining a &amp;quot;voice&amp;quot; register, using the object instead of the subject register if its value is &amp;quot;passive&amp;quot;. But this would require us to assign a value to the voice register in parsing before the relevant evidence is in.</Paragraph>
    <Paragraph position="2"> It would work only if the contents of the voice register were changed at the same time as the passive verb was recognized and the contents of the subject register were moved to the object register. It could indeed be made to work, but the solution is unsatisfactory because it does not reflect any general principle that carries over to other cases. More important, it violates a princple that must be regarded as fundamental for the achievement of reversibility in general, namely that each elementary operation that an arc in the grammar can specify must have two systematically related interpretations, for use in generation and parsing respectively.</Paragraph>
    <Paragraph position="3"> Another solution would be to assign the first noun phrase to a neutral register when it is encountered in parsing, and only to copy it into the subject or object registers when it was finally established which one it belonged in. This neutral register would have to be reflected directly in the structure assigned to the sentence because it would be from there that the first noun phrase in the sentence would have to be taken by the generator. One advantage of this scheme is that a passive marker ~ ,Id no longer be required in the structure of&amp;quot; passive sentences. Instead, the voice of a sentence would be determined by the generator on the asis o~ which register--subject or object--had the same contents as the special neutral register.</Paragraph>
    <Paragraph position="4"> The general principle behind this strategy is that ~he contents of a regi ster are never changed in the course ~f either generation 14.</Paragraph>
    <Paragraph position="5"> or parsing. This is the solution I advocate.</Paragraph>
    <Paragraph position="6"> A name is needed for the neutral register, and topic, or theme, suggest themselves immediately. But consider the case of cleft sentences like &amp;quot;It was Brutus that killed Caesar&amp;quot; and &amp;quot;It was Caesar that Brutus killed&amp;quot; and assume, for the sake of the argument, that these are to be handled as main clauses and not by the relative-clause mechanism. Once again, the underlying grammatical function of the first noun phrase is not known when the parser first encounters it. The problem can be solved by the same device, but of all the names one might choose for the neutral register, &amp;quot;topic&amp;quot; is least appropriate in this instance. Something like focus or comment would be more to the point. What, then, of datives? Consider &amp;quot;He gave Fido a bone&amp;quot; and &amp;quot;He gave Fido to Mary&amp;quot;. The problem here is the noun phrase following the verb. In neither case can we convincingly argue that it is either the topic or the focus of the sentence.</Paragraph>
  </Section>
class="xml-element"></Paper>