Natural Language Query to SQL: Dissecting the semantic word dependency graph

Abstract

Data and Information is the core of Science and Technology. The more we are progressing the more data we are contributing to this vault which will be accessed by millions of people for further learning and uses. Now, the problem is, they are not as versatile as we expect them to be. The amount of information accessible currently is way past the wild estimate of a couple of years prior. No wonder that the digital data is stored in Computational Databases. These data are stored, accessed, updated and retrieved using a complex programming language which is in most of the situation SQL(Structured Query Language) for efficiency and it’s applications. The problem is that we need to learn specific SQL languages to access those data even if it’s for searching or researching purposes. This is the hardest part. We’re creating a huge gap between the data and regular human. What if we don't need to learn any database query languages anymore to retrieve data from databases? We could create an intermediate medium that would capture the natural language query from human, generate the relevant SQL query and retrieve the expected data. This is a continuation of another research where this report has itemized understanding on a different kind of approach where we’d breakdown the words, classify it’s internal characteristics and relationships with other sub contexts and iterate over it’s dependency graph to determine the selection clauses and column matching. Also, we will have a good understanding of what have just been done in this particular theme but focusing on the aggregation queries.