MATE, a Meta Layer Between Natural Language and Database

Nowadays, the knowledge of query languages like SQL is mandatory for accessing the most relevant information concerning a company’s business, stored in richly structured databases. The advancement of Natural Language Processing and Deep Learning research has made it possible to develop different models for the conversion of natural language questions into formalized queries. Although the performance of these models is very satisfactory on internationally established benchmarks for the English language, it is undoubtedly necessary to investigate their portability with respect to the types of databases and natural languages used for formulating the questions. For this reason, we realised Mate (Meta lAyer between naTural language and databasE), a framework for interfacing humans and databases, thus facilitating the access in natural language to the data stored in databases. Indeed, Mate is developed with the aim of accessing information in a very simple way, from the perspective of a human-centered AI.