Auto-conversion from Natural Language to Structured Query Language using Neural Networks Embedded with Pre-training and Fine-tuning Mechanism

From Natural Language to Structured Query Language(NL2SQL) represents an interesting yet challenging process in translating natural questions into database queries in human-machine dialogue systems. NL2SQL requires a high level of semantic understanding, which is difficult to achieve with traditional methods. Although there is a variety of ways to address this problem, most of them do not work satisfactorily. In this work, we propose an approach for NL2SQL using artificial neural networks embedded with pre-training and fine-tuning model, where we use XLNet as the encoding layer for the model, this method is able to better capture the hidden semantic information between natural language and table header of dataset. To make more use of the related information, the table header type information is enhanced in our model. In particular we introduce this information into the calculation of probability values at the output layer. The effectiveness of the proposed method is tested with good performance on WikiSql dataset, nearly 81% accuracy is achieved on the dev dataset.