Using Machine Learning to Generate SPARQL Queries from NL for NoSQL Database

Abstract

SPARQL (SPARQL Protocol and RDF Query Language) is a RDF (Resource Description Framework) query language for “key-value” databases such as NoSQL databases e.g. MongoDB. In the recent times, SPARQL/RDF has emerged into a powerful way of extracting data from database columns that contain multiple values for the same key, and especially wherever the columns are joinable variables in the query. However writing a SPARQL/RDF query is in itself a challenging task since SPARQL/RDF’ syntax is of descriptive and declarative nature and requires technical skills.A problem in applying SPARQL/RDF queries to NoSQL databases is the extraction of irrelevant results without sensing context of given piece of input against a single query. In such cases, a user cannot get the required information from linked data in the same context as wanted. To address such problem, there is need of an intelligent approach that sense the context of given piece of natural language input and generate context oriented output to end user in the form of a SPARQL/RDF query. In this paper, a natural language query based framework is presented to generate SPARQL/RDF queries for NoSQL databases. Here, a challenging task is to map natural language queries to map to SPARQL queries. Our framework accepts input natural language query in English from user and then translates the NL queries into SPARQL/RDF queries. The results of the experiments with the deigned framework reflect importance of such approach used for automated generation of SPARQL/RDF queries for NoSQL databases.