Towards an interface for translating natural language questions to SQL: a conceptual framework from a systematic review

摘要

Querying relational databases using human language is a complex process that must take into consideration the lack of knowledge of SQL by the end users who use the content of these systems. Many studies published in the last decade are based on limited, unbalanced or even small datasets, thing that is by far not realistic for training and evaluating a model. The SQL structure is rich and the scale of complexity is very large. Moreover, the difference between the natural language, which is a language of communication, and the SQL language that contains details of implementation specific to relational databases, adds more complications to the problem. In this paper, we present a full systematic review of the existing approaches in the task of text to SQL and we introduce the details of our suggested architecture as well as our new approach to tackle the problem of nested queries.