<?xml version="1.0" standalone="yes"?>
<Paper uid="P01-1061">
  <Title>Computational properties of environment-based disambiguation</Title>
  <Section position="2" start_page="0" end_page="0" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> Shallow semantic processing applications, comparing argument structures to search patterns or filling in simple templates, can achieve respectable results using the standard 'pipeline' approach to semantics, in which sentences are morphologically and syntactically resolved to a single tree before being interpreted. Putting disambiguation ahead of semantic evaluation is reasonable in these applications because they are primarily run on content like newspaper text or dictated speech, where no machine-readable contextual information is readily available to provide semantic guidance for disambiguation.</Paragraph>
    <Paragraph position="1"> This single-tree semantic architecture is a poor fit for applications such as natural language interfaces however, in which a large amount of contextual information is available in the form of the objects and events in the application's run-time environment. This is because the environment information cannot be used to inform parsing and disambiguation decisions unless the input sentence is semantically analyzed, but this does not occur until after parsing in the single-tree architecture.</Paragraph>
    <Paragraph position="2"> Assuming that no current statistical disambiguation technique is so accurate that it could not benefit from this kind of environment-based information (if available), then it is important that the semantic analysis in an interface architecture be efficiently performed during parsing.</Paragraph>
    <Paragraph position="3"> This paper describes the computational properties of one such architecture, embedded within a system for giving various kinds of conditional instructions and behavioral constraints to virtual human agents in a 3-D simulated environment (Bindiganavale et al., 2000). In one application of this system, users direct simulated maintenance personnel to repair a jet engine, in order to ensure that the maintenance procedures do not risk the safety of the people performing them. Since it is expected to process a broad range of maintenance instructions, the parser is run on a large subset of the Xtag English grammar (XTAG Research Group, 1998), which has been annotated with lexical semantic classes (Kipper et al., 2000) associated with the objects, states, and processes in the maintenance simulation. Since the grammar has several thousand lexical entries, the parser is exposed to considerable lexical and structural ambiguity as a matter of course.</Paragraph>
    <Paragraph position="4"> The environment-based disambiguation architecture described in this paper has much in common with very early environment-based approaches, such as those described by Winograd (Winograd, 1972), in that it uses the actual entities in an environment database to resolve ambiguity in the input. This research explores two extensions to the basic approach however: 1. It incorporates ideas from type theory to represent a broad range of linguistic phenomena in a manner for which their extensions or potential referents in the environment are well-defined in every case. This is elaborated in Section 2.</Paragraph>
    <Paragraph position="5"> 2. It adapts the concept of structure sharing, taken from the study of parsing, not only to translate the many possible interpretations of ambiguous sentences into shared logical expressions, but also to evaluate these sets of potential referents, over all possible interpretations, in polynomial time. This is elaborated in Section 3.</Paragraph>
    <Paragraph position="6"> Taken together, these extensions allow interfaced systems to evaluate a broad range of natural language inputs - including those containing NP/VP attachment ambiguity and verb sense ambiguity - in a principled way, simply based on the objects and events in the systems' environments.</Paragraph>
    <Paragraph position="7"> For example, such a system would be able to correctly answer 'Did someone stop the test at 3:00?' and resolve the ambiguity in the attachment of 'at 3:00' just from the fact that there aren't any 3:00 tests in the environment, only an event where one stops at 3:00.1 Because it evaluates instructions before attempting to choose a single interpretation, the interpreter can avoid getting 'stranded' by disambiguation errors in earlier phases of analysis. null The main challenge of this approach is that it requires the efficient calculation of the set of objects, states, or processes in the environment that each possible sub-derivation of an input sentence could refer to. A semantic interpreter could always be run on an (exponential) enumerated set of possible parse trees as a post-process, to filter out those interpretations which have no environment referents, but recomputing the potential environment referents for every tree would require an enormous amount of time (particularly for broad coverage grammars such as the one employed here). The primary result of this paper is therefore a method of containing the time complexity of these calculations to lie within the complexity of parsing (i.e. within a0a2a1a4a3a6a5a8a7 for a context-free grammar, where a3 is the number of words 1It is important to make a distinction between this environment information, which just describes the set of objects and events that exist in the interfaced application, and what is often called domain information, which describes (usually via hand-written rules) the kinds of objects and events can exist in the interfaced application. The former comes for free with the application, while the latter can be very expensive to create and port between domains.</Paragraph>
    <Paragraph position="8"> in the input sentence), without sacrificing logical correctness, in order to make environment-based interpretation tractable for interactive applications. null</Paragraph>
  </Section>
class="xml-element"></Paper>