Natural Language To NoSQL Query Conversion using Deep Learning

Abstract

NoSQL is a database design strategy capable of accommodating a broad range of data models. NoSQL, which stands for not only SQL, an alternative to traditional relational databases where information is put in tables and information schemes are thoroughly constructed before the database is constructed. NoSQL databases are particularly helpful when working with big information sets. There are different types of NoSQL databases such as Key value data stores, Document stores, Wide column stores and Graph stores. This paper proposes the conversion of English query to the Document stores (Here MongoDB) query. Benefits of NoSQL queries are scalability, performance, high availability, global availability, flexibility, reliability and data modeling. Encoder-Decoder based machine translation method is used for the English to NoSQL query conversion. Encoder neural network is used for creating the Thought Vector of the given English query. Thought vector is the encapsulated form of the given Natural Language Query. Decoder neural network is used for predicting the NoSQL query based on this Thought vector. Teacher forcing method is used for the target sequence prediction purpose. The proposed system handles ten types of MongoDB queries using ten separate deep learning models and shows satisfactory results.

Keywords: Natural Language Processing, Deep Learning, NoSQL, Encoder-Decoder Model, Teacher Forcing