Natural language interfaces to databases: MIS impact, and a survey of their use and importance

• Authors: Vijay Sethi
• Venue: SIGCPR '86 Proceedings of the twenty-second annual computer personnel research conference on Computer personnel research conference
• Year: 1986

The introduction of Artificial Intelligence (AI) tools and applications has led to a growing body of literature analyzing their role in business (Reitman, 1984; International Management (1), 1984; Blanning, 1984; Ford, 1985; Harris, 1985). However, there has been a limited discussion of NL processing in the Management Information Systems (MIS) literature besides a theoretical analysis of the implications of different language views for the design of information systems (Lyytinen, 1985). Though the Computer Science, Psychology, and Computational Linguistics disciplines have developed this field to a great extent, and extensive state-of-the-art summarizations of different NL applications (Slocum, 1985) are useful for providing an overview to managers, most studies do not analyze in detail the dimensions of impact on organizations and decision makers. Currently, when NL based systems are moving out of research laboratories, there is a need for MIS practitioners to understand their benefits and evaluate them for organizational applications.This paper integrates diverse literature on NL interfaces, most of which emphasizes theoretical and technical issues (for instance, the knowledge structures or the parsing strategy that should be used by systems). It develops in detail the dimensions of impact of NL interfaces, an aspect insufficiently discussed by previous studies. It also formalizes one of the only categorizations of NL interfaces developed in the literature. The empirical results could be especially useful for developing implementation strategies for NL based systems.This paper discusses in detail one type of NL systems: NL interfaces to databases (abbreviated to NL interfaces in future). After discussing the reasons for the need of NL interfaces (Section 1), the paper outlines various theoretical capabilities of systems (Section 2). Interfaces are categorized into four types depending on the extent to which they exhibit these capabilities (Section 3). Implications for MIS, in terms of contingencies under which these systems are effective, are then proposed (Section 4). The paper also reports a survey where MIS managers rated the relative importance of various NL applications, the importance of certain capabilities of NL interfaces, and the benefits of such interfaces (Section 5). Following a discussion of the results (Section 6), the conclusion (Section 7) summarizes the paper and highlights the major results.