<?xml version="1.0" standalone="yes"?>
<Paper uid="E87-1037">
  <Title>A Comparison of Rule-Invocation Strategies in Context-Free Chart Parsing</Title>
  <Section position="2" start_page="0" end_page="226" type="intro">
    <SectionTitle>
1 Background
</SectionTitle>
    <Paragraph position="0"> and Introduction An apparent tendency in computational linguistics during the last few years has been towards declarative grammar formalisms. This tendency has manifested itself with respect to linguistic tools, perhaps seen most clearly in the evolution from ATNs with their strongly procedural grammars to PATR-II in its various incarnations (Shieber et al. 1983, Karttunen 1986), and to logic-based formalisms such as DCG (Pereira and Warren 1980). It has also manifested itself in linguistic theor/es, where there has been a development from systems employing sequential derivations in the analysis of sentence structures to systems like LFG and GPSG which establish relations among the elements of a sentence in an order-independent and also direction-independent way. For example, phenomena such as rule ordering simply do not arise in these theories.</Paragraph>
    <Paragraph position="1"> This research has been supported by the National Swedish Board for Technical Development.</Paragraph>
    <Paragraph position="2"> In addition, declarative formalisms are, in principle, processor-independent. Procedural formalisms, although possibly highly standardized (like Woods' ATN formalism), typically make references to an (abstract) machine.</Paragraph>
    <Paragraph position="3"> By virtue of this, it is possible for grammar writers to concentrate on linguistic issues, leaving aside questions of how to express their descriptions in a way which provides for efficient execution by the processor at hand.</Paragraph>
    <Paragraph position="4"> Processing efficiency instead becomes an issue for the designer of the processor, who has to find an overall aoptimal~ control strategy for the processing of the grammar. In particular (and also because of the potentially very large number of rules in realistic natural-language systems), this brings the rule-invocation strategy I into critical focus: to gain maximal processing efficiency, one has to determine the best way of putting the rules to use. 2 This paper focuses on rule-invocation strategies from the perspective of (context-free) chart parsing (Kay 1973, 1982; Kaplan 1973).</Paragraph>
    <Paragraph position="5"> Context-free phrase-structure grammar is of interest here in particular because it is utilized as the backbone of many declarative formalisms. The chart-parsing framework is of interest in this connection because, being a C'higher-order algorithm&amp;quot; (Kay 1982:329), it lends itself easily to the processing of different grammatical formalisms. At the same time it is of course a natural test bed for experiments with various control strategies.</Paragraph>
    <Paragraph position="6"> Previously a number of comparisons of rule-invocation strategies in this or in similar settings have been reported: ZThis term seems to have been coined by Thompson (1981). Basically, it refers to the spectrum between top-down and bottom-up processing of the grammar rules.</Paragraph>
    <Paragraph position="7"> 2The other principal control-strategy dimension, the search ~g;/(depth-first vs. breadth-first), is irrelevant for the efficiency in chart parsing since it only affects the order in which successive (partial) analyses are developed.</Paragraph>
    <Paragraph position="8">  Kay (1982) is the principal source, providing a very general exposition of the control strategies and data structures involved in chart parsing. In considering the efficiency question, Kay favours a ~directed ~ bottom-up strategy (cf. section 2.2.3}.</Paragraph>
    <Paragraph position="9"> Thompson (1981) is another fundamental source, though he discusses the effects of various rule-invocation strategies mainly from the perspective of GPSG parsing which is not the main point here.</Paragraph>
    <Paragraph position="10"> Kilbury (1985) presents a left-corner strategy, arguing that with respect to natural-language grammars it will generally outperform the top-down (Earley-style) strategy.</Paragraph>
    <Paragraph position="11"> Wang (1985) discusses Kilbury's and Earley's algorithms, favouring the latter because of the inefficient way in which bottom-up algorithms deal with rules with right common factors. Neither Wang nor Kilbury considers .the natural approach to overcoming this problem, viz. top-down filtering (of. section 2.2.3).</Paragraph>
    <Paragraph position="12"> As for empirical studies, Slocum (1981) is a rich source. Among many other things, he provides some performance data regarding top-down filtering.</Paragraph>
    <Paragraph position="13"> Pratt (1975) reports on a successful augmentation of a bottom-up chart-like parser with a top-down filter.</Paragraph>
    <Paragraph position="14"> Tomita (1985, 1986) introduces a very efficient, extended LR-parsing algorithm that can deal with full context-free languages. Based on empirical comparisons, Tomita shows his algorithm to be superior to Earley's algorithm and also to a modified version thereof (corresponding here to %elective topdownS; cf. section 2.1.2). Thus, with respect to raw efficiency, it seems clear that Tomita's algorithm is superior to comparable chart-parsing algorithms.</Paragraph>
    <Paragraph position="15"> However, a chart-parsing framework does have its advantages, particularly in its flexibility and openendedness. null The contribution this paper makes is: to survey fundamental strategies for rule-invocation within a context-free chart-parsing framework; in particular to specify ~directed ~ versions of Kilbury's strategy; and * to provide a practical comparison of the strategies based on empirical results.</Paragraph>
  </Section>
class="xml-element"></Paper>