Showing papers on "Natural language understanding published in 1990"

A Probabilistic Approach to Language Understanding

Abstract: We discuss a new framework for text understanding. Three major design decisions characterize this approach. First, we take the problem of text understanding to be a particular case of the general problem of abductive inference: reasoning from effects to causes. Second, we use probability theory to handle the uncertainty which arises in abductive inference in general, and natural language understanding in particular. Finally, we treat all aspects of the text understanding problem in a unified way. All aspects of natural language processing are treated in the same framework, allowing us to integrate syntactic, semantic and pragmatic constraints. In order to apply probability theory to this problem, we have developed a probabilistic model of text understanding. To make it practical to use this model, we have devised a way of incrementally constructing and evaluating belief networks that is applicable to other abduction problems. We have written a program, Wimp3, to experiment with this framework.

Learning from textbook knowledge: a case study

Abstract: One of the "grand challenges for machine learning" is the problem of learning from textbooks. This paper addresses the problem of learning from texts including omissions and inconsistencies that are clarified by illustrative examples. To avoid problems in natural language understanding, we consider a simplification of this problem in which the text has been manually translated into a logical theory. This learning problem is solvable by a technique that we call analogical abductive explanation based learning (ANA-EBL). Formal evidence and experimental results in the domain of contract bridge show that the learning technique is both efficient and effective.

Evaluating computer-generated domain-oriented vocabularies

Abstract: It is generally accepted that natural language understanding systems are not now able to deal successfully with unrestricted text, except in very superficial ways. Certainly no current NL system exhibits any significant degree of understanding over arbitrary subject matter. Moreover, there is no convincing reason to believe this situation will change in the near future. Successful systems, therefore, have been restricted to specific applications in particular discourse domains. In those situations where users are expected to provide the domain vocabulary (e.g., TEAM, TQA, etc.) it would be very desirable to provide at least suggestions as to what this vocabulary might be, because a good part of the difficulty in customizing a general system consists of supplying the domain vocabulary and specifying its grammatical properties. This paper discusses some methods for identifying domain vocabulary, as well as techniques for evaluating the quality of the resulting word list.

Knowledge representation in LILOG

Abstract: This paper introduces the knowledge representation language L LILOG . The language is being developed in the framework of the LILOG project and serves for modelling the semantic background knowledge of the LILOG natural language understanding system. Moreover, it is also used as the target language for representing information extracted from German texts in a logical form. The aspects of L LILOG discussed here focus on the sort concept of L LILOG and its means for structuring knowledge bases. The sort concept of L LILOG integrates ideas from the KL-ONE family of languages and other feature term languages having their origin in the area of computational linguistics into the framework of an order-sorted predicate logic. The structuring concept introduced for L LILOG is a simple form of separating logical theories into modules.

Natural language understanding and speech recognition

Abstract: Natural language understanding must be an integral part of any automatic speech recognition system that attempts to deal with interactive problem solving. The methods for representing and integrating knowledge from different sources may be valuable for the understanding process as well as speech recognition.

CARAMEL: a flexible model for interaction between the cognitive processes underlying natural language understanding

Abstract: In this paper we present a general natural language processing system called CARAMEL (in French: Comprehension Automatique de Recits, Apprentissage et Modelisation des Echanges Langagiers). Over the last few years our group has developed many programs to deal with different aspects of natural language processing. This paper describes a general architecture that integrates them in a flexible way, and provides a control strategy capable of adapting itself to the requirements of a particular task. The model is composed of three fundamental elements:- a structured memory containing permanent knowledge and working structures of the system- a set of processes, dedicated to the execution of the various cognitive tasks- a supervisor, whose function is to trigger, to run coherently and to synchronize the processes.The system contains a kind of blackboard, which is enhanced with a control mechanism driven by meta-rules. This architecture is fully implemented. We are currently developing the meta-rules necessary to use the model for various tasks.

Towards understanding text with a very large vocabulary

Abstract: In order to meet the information processing demands of the next decade, natural language systems must have the capability of processing very large amounts of text, commonly called "messages", from highly diverse sources written in any of a few dozen languages. One of the key issues in building systems with this scale of competence is handling large numbers of different words and word senses. Natural language understanding systems today are typically limited to vocabularies of less than 10,000 words; tomorrow's systems will need vocabularies at least 5 times that to effectively handle the volume and diversity of messages needing to be processed.

A Probabilistic Approach to Text Understanding

Abstract: We discuss a new framework for text understanding. Three major design decisions characterize this approach. First, we take the problem of text understanding to be a particular case of the general problem of abductive inference: reasoning from effects to causes. Second, we use probability theory to handle the uncertainty that arises in abductive inference in general and natural language understanding in particular. Finally, we treat all aspects of the text-understanding problem in a unified way. All aspects of natural language processing are treated in the same framework, allowing us to integrate syntactic, semantic and pragmatic constraints. In order to apply probability theory to this problem, we have developed a probabilistic model of text understanding. To make it practical to use this model, we have devised a way of incrementally constructing and evaluating belief networks that is applicable to other abduction problems. We have written a program, Wimp3, to experiment with this framework.

Eventualities in a natural language understanding system

Abstract: This paper focuses on the discussion of suitable representations of eventualities in a formal language and of the possibilities to draw inferences from representations. We argue in favour of the treatment of eventualities as individuals, which are structured along different lines. When reifying eventualities, there are different possibilities of individualization. This is similarily true for the domain of objects. Thus we investigate these possibilities in parallel with objects, and obtain a rather symmetric structuring of the domain of individuals, i.e. a sort hierarchy which is sensible for different kinds of eventualities and objects respectively.

Dehumanized people and humanized programs: a natural language view of being there

Abstract: The language used by Chance the gardener, the main character in Jerzy Kosinski's Being There, is uncannily like that of several well known natural language understanding systems: ELIZA, PARRY and SAM. This natural language understanding (NLU) view of Chance meshes remarkably well with literary criticism of the character and the book. In light of these resemblances, there is some (not-too-serious) discussion of how NLU systems might perform more intelligently, or at least might appear more intelligent.

Solving ambiguities in the semantic representation of texts

Abstract: One of the issues of Artificial Intelligence is the transfer of he knowledge conveyed by Natural Language into formalisms that a computer can interpret. In the Natural Language Processing department of the IBM France Paris Scientific Center, we are developing and evaluating a system prototype whose purpose is to build a semantic representation of written French texts in a rigorous formal model (the Conceptual Graph model, introduced by J. F. Sowa[10]).The semantic representation of texts may then be used in various applications, such as intelligent information retrieval. The accuracy of the semantic representation is therefore crucial in order to obtain valid results in any subsequent applications. In this article we explain how ambiguities related to Natural Language may be solved by semantic analysis using the Conceptual Graph model.

A Measure of Semantic Relatedness for Resolving Ambiguities in Natural Language Database Requests

Abstract: A measure of semantic relatedness based on distance between objects in the database schema has previously been used as a basis for solving a variety of natural language understanding problems including word sense disambiguation, resolution of semantic ambiguities, and attachment of post noun modifiers. The use of min/max values which are usually recorded as part of the process of designing the database schema is proposed as a basis for solving the given problems as they arise in natural language database requests. The min/max values provide a new source of knowledge for resolving ambiguities and a semantics for understanding what knowledge has previously been used by distance measures in database schemas.

A mechanism for ellipsis resolution in dialogued systems

Abstract: An ellipsi resolution mechanism is presented. The mechanism is a part of a Natural Language Understanding System developed in the last years in order to be included as a main component of several projects based on man/machine interactions. CAPRA, an intelligent system for teaching programming, and GUAI, a natural language interfaces generator, are two of such applications. In our approach, syntactic and knowledge-based techniques are combined in order to get a great coverage of elliptical cases.

A formal semantics for internal localization: an essay on spatial commonsense knowledge

Abstract: The definition of a formal semantics for spatial structures whose expression Natural Language allows is an important contribution to Natural Language Understanding as well as to reasoning automation. This paper presents a semantics for the Internal Localization Nouns and Adjectives (ILN, ILA) as avant (front), haut (top), devant (front extremity), etc, in a first order language. In fact, to get those results, some fundamental concepts and relations must be first defined from studies in linguistics and cognitive psychology. We introduce a spatial ontology which import goes beyond the ILN's and ILA's alone. At last, from their semantics, we present some models for spatial reasoning.

The Understanding Algorithms

Abstract: The final goal of a continuous speech understanding system is the generation of a representation of the utterance meaning, beside the recognition of the utterance words. From this representation a proper action can be taken in order to satisfy the needs of the user that interacts with the system (for instance by giving him an answer to a question). Both activities, word recognition and understanding, have to be performed and should take advantage of available knowledge about words, language and domain. Recognition must use that knowledge as a source of constraints for word disambiguation while the understanding activity is entirely based on that knowledge and requires the same effort as in the case of written natural language understanding.

Statistical techniques for parsing messages

Abstract: Message processing is the extraction of information about key events described in brief narratives concerning a narrow domain. This is a suitable task for natural language understanding, since the amount of world knowledge required is limited. However, the messages are often ill-formed and therefore require the grammar which parses them to be quite forgiving. This often results in a proliferation of parses. This problem is compounded by one's inability to construct a complete domain model which would resolve all the semantic ambiguity. Thus, selection of the correct parse becomes an important goal for such systems. Structural preference is a technique which helps disambiguation by assigning a higher preference to certain syntactic structures. The idea of statistical parsing evolved from the desire of being able to prefer certain structures over others on the basis of empirical observations, rather than ad-hoc judgement. In the framework of statistical parsing, every production of the grammar is assigned a priority, which is computed from a statistical analysis of a corpus. There are two distinct methodologies that can be used for assigning these priorities. In Supervised Training, only the correct parses are used for training the grammar. On the other hand, Unsupervised Training uses parses independent of their semantic validity. After assigning the priorities, the parser searches for parses in a best-first order as dictated by these priorities. When this scheme was incorporated into the PROTEUS message understanding system while processing OPREP (U.S. Navy Operational) messages, a two-fold advantage was observed. Firstly, the speed of the parsing increased, because rare productions tended not to get used at all. Secondly, since the parses were generated in the best-first order, the parses generated earlier on tended to be more likely and semantically more acceptable. The performance of the modified parsing algorithm was evaluated with and without several refinements such as the use of context sensitive statistics and the use of heuristic penalties. The relative performances of the grammars trained by Supervised Training and Unsupervised Training were also compared.

Integrating analogical reasoning in a natural language understander

Abstract: The research described in this paper addresses the problem of integrating analogical reasoning and argumentation into a natural language understanding system. We present an approach to completing an implicit argument-by-analogy as found in a natural language editorial text. The transformation of concepts from one domain to another, which is inherent in this task, is a complex process requiring basic reasoning skills and domain knowledge, as well as an understanding of the structure and use of both analogies and arguments. The integration of knowledge about natural language understanding, argumentation, and analogical reasoning is demonstrated in a proof of concept system called ARIEL. ARIEL is able to detect the presence of an analogy in an editorial text, identify the source and target components, and develop a conceptual representation of the completed analogy in memory. The design of our system is modular in nature, permitting extensions to the existing knowledge base and making the argumentation and analogical reasoning components portable to other understanding systems.

Volume 3: A Guide to IRUS-II Application Development

Abstract: (Cawmnw anl mww" it inqcenavy "W a-clbI b10y r=tma IRUS-II is the understanding subsystem of the Janus natural language interface. IRUS-11 isa natural language understanding (NLU) shell. That is, it contains domain-independentalgorithms, a large graimmar of English, domait-n-independent semantic interpretation rules,and a domain-independent discourse component. In addition, several software aids areprovided to customize the system to particular application domains. These software aidsoutput the four knowledge bases necessary for IRUS-II to correctly interpret Englishutterances and generate appropriate code for simultaneous access to multiple applicationsystems.This document describes the four knowledge bases and how to construct them.This is the third volume of a three volume final report. "0. OITRIlUTIONirtvAiL&ugUry OF AsSTRACr 2:t. AaSTRACT SICU~iTY CLASSIFICATIONSUECLAS$IFIE~oALmItoifo 0 SAMEC AS APT US In * MEi Of StEPONiSIiL -1,0VIiOUA. al. TELEPI4ONIE (kWnCM