Natural Language Systems Research Articles

An English language question answering system for a large relational database

By typing requests in English, casual users will be able to obtain explicit answers from a large relational database of aircraft flight and maintenance data using a system called PLANES. The design and implementation of this system is described and illustrated with detailed examples of the operation of system components and examples of overall system operation. The language processing portion of the system uses a number of augmented transition networks, each of which matches phrases with a specific meaning, along with context registers (history keepers) and concept case frames; these are used for judging meaningfulness of questions, generating dialogue for clarifying partially understood questions, and resolving ellipsis and pronoun reference problems. Other system components construct a formal query for the relational database, and optimize the order of searching relations. Methods are discussed for handling vague or complex questions and for providing browsing ability. Also included are discussions of important issues in programming natural language systems for limited domains, and the relationship of this system to others.

Developing a natural language interface to complex data

Aspects of an intelligent interface that provides natural language access to a large body of data distributed over a computer network are described. The overall system architecture is presented, showing how a user is buffered from the actual database management systems (DBMSs) by three layers of insulating components. These layers operate in series to convert natural language queries into calls to DBMSs at remote sites. Attention is then focused on the first of the insulating components, the natural language system. A pragmatic approach to language access that has proved useful for building interfaces to databases is described and illustrated by examples. Special language features that increase system usability, such as spelling correction, processing of incomplete inputs, and run-time system personalization, are also discussed. The language system is contrasted with other work in applied natural language processing, and the system's limitations are analyzed.

HAM-RPM

During the past two years we designed and implemented an understanding system named HAM-RPM, which belongs to a class of natural language systems commonly referred to as parsed, contentmotivated (Wilks) or second generation systems (Winograd). The current system is operating within a question-answering environment and allows the complete comprehension process from the understanding of the German input to the generation of a conversationally appropriate response to take place.

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.

Computer acquisition of natural language: a review and prospectus

A computer program which could learn to converse in natural language, in a fashion analogous to the child's acquisition of his mother tongue, would benefit theories of language acquisition and efforts toward mechanical language understanding. This paper opens with a sketch of the acquisition process as presently understood by linguists, and a discussion of models and methods. The “state of the art” in computer acquisition of natural language is then considered, in a critical review of four computer oriented natural language systems (Schwarcz, 1967; Kelley, 1967; Harris, 1972; Block et al ., 1975) with important acquisition components. Based in large part on what has been learned from these previous efforts, a new Comprehensive Language Acquisition Program (CLAP) is proposed. The heart of CLAP is the development of its parsing and related components through the sequential activation of five strategies: segmentation and meaning association, linear ordering, structural generalization, conflict resolution, and using discourse. The first three of these strategies are sufficiently well defined for computer implementation, using established methodologies from artificial intelligence.

Stemming as a system design consideration

A good deal of work has been done over the years trying to demonstrate the feasibility or superiority of automatic indexing techniques. (1--8) Most of these techniques are designed for natural language systems, i. e., systems which have an uncontrolled vocabulary. While they incorporate many desirable features, they also have some disadvantages.

Natural language storage and retrieval of medical diagnostic information: Experience at the UCLA Hospital and Clinics over a 10-year period

A natural language storage and retrieval system initially designed for pathology reports has been in operation for ten years at UCLA Hospital and Clinics. The original system has been improved to provide a thesaurus processor with added capabilities for expanding search request terms and a newly developed set of search programs with user options that make complex and more accurate retrievals possible. Summarized diagnostic statements or impressions from five specialties (Surgical Pathology, Bone Marrows, Autopsy, Nuclear Medicine, and Neuroradiology), are automatically encoded by referencing a master computer dictionary containing a unique numeric code for each English word. Input and retrievals are batch, off-line operations in free text, and are currently processed on the IBM 370/145. User acceptance is high, quality of retrievals improved, and cost is nominal. This paper describes an expanded natural language system which is now providing a viable form of information storage and retrieval for research, medical education, patient care, quality control, statistics and administrative purposes.

Interactive consulting via natural language

Interactive programming systems often contain help commands to give the programmer on-line instruction regarding the use of the various systems commands. It is argued that it would be relatively easy to make these help commands significantly more helpful by having them accept requests in natural language. As a demonstration, Weizenbaum's ELIZA program has been provided with a script that turns it into a natural language system consultant.

On the Need for a Phonological Base

Pre-Chomskyan linguistics recognized that the phonological systems of natural languages tend to form symmetrical patterns; this insight could be incorporated into generative phonology via a phonological base component, but is not adequately expressed by MS rules. A phonological base would also solve some difficulties in the theory of systematic phonetics. Current lack of interest in phonological patterning has resulted from the increased attention paid to morphophonemics; but the deep vs. surface distinction in phonology can be demonstrated on the basis of patterning even in languages lacking morphophonemic alternation. That distinction, it is argued, may be of only historical validity, whether or not alternations occur in the language.'

MANIPULATION OF AUTOPSY DIAGNOSES BY COMPUTER TECHNIQUE.

A new system of coding and storing autopsy records for searching by computer has been tested on a sample of 500 records. Coding is done by computer routines, handcoding being eliminated. Records are put into the system in natural language, as they are originally written. The separate words are translated into a classification code based principally on pathological process and anatomic site. All information is coded; nothing is deleted. Twenty-seven questions representing a wide range of information requests (generic to specific) were searched. The results (3,115 total answers of which three were spurious) indicated high accuracy for the system. Possible widespread application is suggested.