THE TEMPLATE-BASED METHOD FOR DESIGNING A RELATIONAL DATABASE CONSTRUCTOR
Kateryna Yalova, Mykhailo Babenko, Andrii Zhuravel · Collection of scholarly papers of Dniprovsk State Technical University (Technical Sciences) · 2024
The concept of using natural language instead of SQL commands laid the foundation for the development of the Natural Language Interface to Database (NLIDB) approach. Despite significant progress and a variety of research and applied developments, the task of developing NLIDB remains an ongoing scientific and practical challenge. Issues such as handling ambiguity, understanding context, and the significant variability in natural language command expression impact the quality of text-to-command conversion, necessitating further exploration of effective solutions and mechanisms. Numerous studies have demonstrated the effectiveness of implementing NLIDB (Natural Language Interface to Database) in the creation and management of relational databases (RDBs). The objective of this study is to design an NLIDB software application in the form of an RDB designer, which facilitates a dialogue with the user — an expert in the subject area — in such a way that this interaction results in the design of the RDB architecture. The knowledge provided by the user about the subject area is translated into SQL queries using the «Create» keyword to generate the database and its corresponding tables. To achieve the research objectives, we successfully addressed the tasks of developing the architecture of the RDB designer, creating dialogue templates and a user interaction model, and constructing an algorithm that transforms the verbal description of the subject area into a formalized database architecture. The architecture of the RDB constructor developed using NLIDB consists of two primary modules: a natural language dialogue module and an SQL query generation module. The guided dialogue with the user begins with the input of general data about the subject area and culminates in the output of either a list of business rules, a database diagram, SQL queries, or an .mdf file, depending on the user's requirements. The direction of further research is the use of NLIDB to create a software tool that transforms text messages into SQL queries to extract data from a database based on the Select keyword. The developed database constructor can be used as an educational tool to enhance the process of acquiring practical skills during laboratory classes in the databases course.