NL2SQL method combining semantic analysis and semantic component matching

• Main Authors: DU KUNPENG, ZHANG XUAN, WANG XU, NONG QIONG, GAO CHEN, LIANG YUQIN, WANG JISHU, LI LINYU, MA YUBIN
• Format: Patent
• Language: Chinese; English
• Published: 29.11.2022
• Subjects: CALCULATING; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; PHYSICS

The invention provides an NL2SQL method combining semantic analysis and semantic component matching, which comprises the following steps: S1, data preprocessing; S2, combining semantic slot filling and aggregation function identification: S21, encoding a natural language query question Quest; s22, predicting an aggregation function type and an operator type based on a result of S21, and predicting a column field at the same time; s23, performing joint modeling on objective functions of the aggregation function type prediction task, the operator type prediction task and the column field prediction task to obtain an aggregation function type value, an operator type and a column field label list; s3, constructing a semantic candidate pair; and S4, semantic component matching. The problems that in the form of a large and wide table, labeling tasks are heavy, an existing method is poor in migration capability, universality is not enough in different enterprise business scenes, a conventional assembly line method i