Ralph M. Weischedel

Living up to expectations: computing expert responses

In cooperative man-machine interaction, it is necessary but not sufficient for a system to respond truthfully and informatively to a users question. In particular, if the system has reason to believe that its planned response might mislead the user, then it must block that conclusion by modifying its response. This paper focusses on identifying and avoiding potentially misleading responses by acknowledging types of informing behavior usually expected of an expert. We attempt to give a formal account of several types of assertions that should be included in response to questions concerning the achievement of some goal (in addition to the simple answer), lest the questioner otherwise be misled.

An artificial intelligence approach to language instruction

This paper describes an implemented, prototype system for a sophisticated, intelligent tutor for instruction in a foreign language. The system is an application of artificial intelligence research in natural language, but it implements several ideas that depart from standard approaches to natural language understanding. For instance, the semantic analyzer diagnoses several kinds of comprehension problems and semantic errors that a student might make. Some fine distinctions in meaning are represented to detect misuse of words. Not only is a model of good syntax included in the tutor, but also a model of incorrect forms, rich enough to pinpoint specific syntactic mistakes. Finding the intended interpretation is complicated by the likelihood of student errors. Therefore, perfect syntactic form is not necessary for semantic analysis of the students input. The problems discussed and solutions presented are closely related to the more general problem of how to respond to a natural language input that surpasses the computers model of language or of context.

Semantic Interpretation Using KL-ONE

This paper presents extensions to the work of Bobrow and Webber [Bobrow & Webber 80a, Bobrow & Webber 80b] on semantic interpretation using KL-ONE to represent knowledge. The approach is based on an extended case frame formalism applicable to all types of phrases, not just clauses. The frames are used to recognize semantically acceptable phrases, identify their structure, and, relate them to their meaning representation through translation rules. Approaches are presented for generating KL-ONE structures as the meaning of a sentence, for capturing semantic generalizations through abstract case frames, and for handling pronouns and relative clauses.

An Improved Heuristic for Ellipsis Processing

Several natural language systems (e.g., Bobrow et al., 1977; Hendrix et al., 1978; Kwasny and Sondheimer, 1979) include heuristics for replacement and repetition ellipsis, but not expansion ellipsis. One general strategy has been to substitute fragments into the analysis of the previous input, e.g., substituting parse trees of the elliptical input into the parse trees of the previous input in LIFER (Hendrix, et al., 1978). This only applies to inputs of the same type, e.g., repeated questions.

A rule-based approach to ill-formed input

Though natural language understanding systems have improved markedly in recent years, they have only begun to consider a major problem of truly natural input: ill-formedness. Quite often natural language input is ill-formed in the sense of being misspelled, ungrammatical, or not entirely meaningful. A requirement for any successful natural language interface must be that the system either intelligently guesses at a users intent, requests direct clarification, or at the very least, accurately identifies the ill-formedness. This paper presents a proposal for the proper treatment of ill-formed input. Our conjecture is that ill-formedness should be treated as rule-based. Violation of the rules of normal processing should be used to signal ill-formedness. Meta-rules modifying the rules of normal processing should be used for error identification and recovery. These meta-rules correspond to types of errors. Evidence for this conjecture is presented as well as some open ~]estions.

If The Parser Fails

The unforgiving nature of natural language components when someone uses an unexpected input has recently been a concern of several projects. For instance, Carbonell (1979) discusses inferring the meaning of new words. Hendrix, e t .a l . (1978) describe a system that provides a means for naive users to define personalized paraphrases and that l i s ts the items expected next at a point where the parser blocks. Weischedel, e t .a l . (1978) show how to relax both syntactic and semantic constraints such that some classes of ungrammatical or semantically inappropriate input are understood. Kwasny aod Sondheimer (1979) present techniques for understanding several classes of syntactically il l-formed input. Codd, e t .a l . (1978) and Lebowitz (1979) present alternatives to top-down, le f t to r igh t parsers as a means of dealing with some of these problems.

PROBLEM LOCALIZATION STRATEGIES FOR PRAMATICS PROCESSING IN NATURAL-LANGUAGE FRONT ENDS

Problem localization is the identification of the most significant failures in the AND-OR tree resulting from an unsuccessful attempt to achieve a goal, for instance, in planning, backward-chaining inference, or top-down parsing. We examine heuristics and strategies for problem localization in the context of using a planner to check for pragmatic failures in natural language input to computer systems, such as a cooperative natural language interface to Unix™ Our heuristics call for selecting the most hopeful branch at ORs, but the most problematic one at ANDs. Surprise scores and special-purpose rules are the main strategies suggested to determine this.

HANDLING ILL-FORMED INPUT: SESSION INTRODUCTION

In natural language access (e.g. English access) to information systems, the magnitude of the problem of absolute ill-formedness can be seen in several case studies. If one includes telegraphic and elliptical constructions in the class of absolute ill-formedness, then case s~udies reported in Thompson (1980) and Eastman and McLean (1981) indicate that as much as 25% of queries to questionanswering systems are absolutely illformed. On the other hand, no matter how large the dictionary, grammar, and underlying system, there will always be unknown words and phrases (e.g. proper names) and impossible requests (due to user misconceptions of the capabilities of the underlying system).

Natural language processing systems and III-formed input: University of Delaware/Sperry Univac

The goal of our project is to add robustness to natural language understanding systems by adding rule-based methods of handling illformed input [1,2,3]. This includes both ungrammatical and semantically inappropriate utterances. The method being used consists of the following procedure:

Natural language processing research at the University of Delaware

A fundamental problem of natural language systems is to aid a user whose input is not understood by the system. Linguistic knowledge and world knowledge must be represented and organized in a way that the system can inform the user of what it did understand and hypothesize reasons for its not understanding the rest.