TalkSQL: A Tool for the Synthesis of SQL Queries from Verbal Specifications

Abstract

Recent advances in the field of Natural Language Processing (NLP) have led to many robust user interfaces (UIs) designed as intelligent tutoring systems (ITS) that help students learn, query and access data in relational databases. Such tools are generally referred to as Natural Language Interfaces to Databases (NLIDBs). Many of these UIs rely on voice or typewritten for further processing. Research has shown that typewritten remains the preferred input method used by database UIs designers for querying relational databases due to its flexibility. Still, there is a dearth of tools that require voice-based inputs for querying relational databases. Despite the scarcity of these tools, many of them fail to provide a comprehensive feedback to a user. In this paper, we introduce a voice-based query system named TalkSQL that takes voice inputs from a user, converts these words into SQL queries and returns a feedback to the user. Automatic feedback generation is of immense importance. To achieve this, we have used regular expressions, a representation of regular languages for the recognition of the Create, Read, Update, Delete (CRUD) operations in SQL and automatically generate a feedback using pre-defined templates. A survey on 53 participants showed that 91.2% agreed that they were able to understand the CRUD command using TalkSQL. The expected contributions are in two-fold: this work may assist a special (e.g. visually impaired) learner to understand SQL queries, and show that a voice-based interface can assist users in understanding SQL queries.