Configurable method and system for converting natural language into sql

Abstract

The invention discloses a configurable method and system for converting a natural language into sql, and the method specifically comprises the steps: constructing an sql structure to carry out the preprocessing of a question text, removing unimportant words, carrying out the discrimination of an aggregation condition through semantic component matching, and determining an aggregation condition element needed in an sql template through semantic component matching judgment; extracting a text time period by utilizing time elements, then obtaining conditions except the time elements through a text, realizing data query and analysis requirements through a natural language interaction mode, providing comprehensive support for complex business logic, and quickly adjusting the system when a new data access questioning method exists; extracting conditions except the time elements through the condition elements, recording each data source and each piece of semantic component information to be matched under the data table, and helping a user to realize data query and analysis requirements through a natural language interaction mode.

Description

technical field [0001] The invention relates to a language processing technology, in particular to a configurable natural language-to-sql conversion method and system. Background technique [0002] At present, the machine natural language to sql method can only support the simple select F(A) from B where C D E structure, where A is the table result field you want to query; F is the function operation on this table field, such as sum, count etc.; B is the table name of the query; C is the field name in the condition; E is the value corresponding to the field in the condition; D is the relationship between the field and the value in the condition, such as greater than or less than. [0003] There are generally two types of element extraction algorithms in this structure: one is to determine the values ​​of A~F through the matching of keywords and regular expressions; the other is to train the training samples through machine learning algorithms, because A~F The value of the F...

AI Technical Summary

Problems solved by technology

[0004] The disadvantages of this mode mainly include the following two points: 1. The sql structure is fixed, and advanced sql generation such as aggregation and table association cannot be performed; 2. The applicable database and business scenarios are single. If a new table is added in the database, or If a new field is added to the table, it needs to be retrained

Images