System, method, and computer program for converting a natural language query to a structured database query
The present disclosure describes a system, method, and computer program for converting a natural language query to a structured database query. In response to receiving a natural language query for a database, an NLU model is applied to the query to identify an intent and entities associated with the query. The intent is mapped to a database object, and candidate query fields and operands are identified from the entities. The candidate query fields and operands are evaluated to identify any subject fields, conditional expressions, record count limit, and ordering/sorting criteria for the query. This including matching certain query fields and operands based on query parameters, operand types, and locations of operands relative to query fields. A query plan is created based on the evaluation of the candidate query fields and operands, and a database query is generated from the query plan.
1. A method, performed by a computer system, for creating a database query from a user's natural language query, the method comprising:
receiving a user's natural language query via a natural language interface to the system;
applying a natural language model to the user's query to identify an intent and a plurality of entities associated with the user's query, wherein the entities are tagged with an entity type and associated with a normalized value;
identifying a database object corresponding to the identified intent;
identifying candidate query fields and operands for the query based on the entities and associated type tags;
creating a query plan for the query by evaluating the candidate query fields and operands to identify subject fields, conditional parameters, any record count limit, and any explicit or implicit ordering criteria for the query, wherein evaluating the candidate query fields and operands comprises:
obtaining query parameters,
determining if any of the candidate query fields are subject fields,
matching the candidate query fields that are not subject fields to operands based on the query parameters, entity tags associated with the operands, and locations of the operands relative to candidate query fields,
matching any operand not matched with a candidate query field to a default query field specified in the query parameters, and
adding the matched candidate queries and operands to the query plan as conditional parameters for the query,
wherein, in response to identifying implicit ordering criteria based on time, the computer system adapts ordering criteria based on the user's query to ensure results that have immediate time significance are placed ahead of less-immediate results; and
creating, based on the query plan, a database query, wherein the database query is in a database query language.
2. The method of claim 1 , further comprising adding a record count limit and an ordering criteria to the query plan.
3. The method of claim 1 , wherein the system identifies any queryable field located between an interrogative or a lookup action entity and an entity corresponding to the database object as a subject field.
4. The method of claim 1 , wherein prior to determining if any of the candidate query fields are subject fields, the system preprocesses the candidate query fields and operands.
5. The method of claim 4 , wherein preprocessing the operands and candidate query fields comprises removing any redundant, trivial, and subsumed query fields and operands from the candidate query fields and operands.
6. A non-transitory computer-readable medium comprising a computer program, that, when executed by a computer system, enables the computer system to perform the following method for creating a database query from a user's natural language query, the method comprising:
receiving a user's natural language query via a natural language interface to the system;
applying a natural language model to the user's query to identify an intent and a plurality of entities associated with the user's query, wherein the entities are tagged with an entity type and associated with a normalized value;
identifying a database object corresponding to the identified intent;
identifying candidate query fields and operands for the query based on the entities and associated type tags;
creating a query plan for the query by evaluating the candidate query fields and operands to identify subject fields, conditional parameters, any record count limit, and any explicit or implicit ordering criteria for the query, wherein evaluating the candidate query fields and operands comprises:
obtaining query parameters,
determining if any of the candidate query fields are subject fields,
matching the candidate query fields that are not subject fields to operands based on the query parameters, entity tags associated with the operands, and locations of the operands relative to candidate query fields,
matching any operand not matched with a candidate query field to a default query field specified in the query parameters, and
adding the matched candidate queries and operands to the query plan as conditional parameters for the query,
wherein, in response to identifying implicit ordering criteria based on time, the computer system adapts ordering criteria based on the user's query to ensure results that have immediate time significance are placed ahead of less-immediate results; and
creating, based on the query plan, a database query, wherein the database query is in a database query language.
7. The non-transitory computer-readable medium of claim 6 , further comprising adding a record count limit and an ordering criteria to the query plan.
8. The non-transitory computer-readable medium of claim 6 , wherein the system identifies any queryable field located between an interrogative or a lookup action entity and an entity corresponding to the database object as a subject field.
9. The non-transitory computer-readable medium of claim 6 , wherein prior to determining if any of the candidate query fields are subject fields, the system preprocesses the candidate query fields and operands.
10. The non-transitory computer-readable medium of claim 9 , wherein preprocessing the operands and query fields comprises removing any redundant, trivial, and subsumed query fields and operands from the candidate query fields and operands.
11. A computer system for creating a database query from a user's natural language query, the system comprising:
one or more processors;
one or more memory units coupled to the one or more processors, wherein the one or more memory units store instructions that, when executed by the one or more processors, cause the system to perform the operations of:
receiving a user's natural language query via a natural language interface to the system;
applying a natural language model to the user's query to identify an intent and a plurality of entities associated with the user's query, wherein the entities are tagged with an entity type and associated with a normalized value;
identifying a database object corresponding to the identified intent;
identifying candidate query fields and operands for the query based on the entities and associated type tags;
creating a query plan for the query by evaluating the candidate query fields and operands to identify subject fields, conditional parameters, any record count limit, and any explicit or implicit ordering criteria for the query, wherein evaluating the candidate query fields and operands comprises:
obtaining query parameters,
determining if any of the candidate query fields are subject fields,
matching the candidate query fields that are not subject fields to operands based on the query parameters, entity tags associated with the operands, and locations of the operands relative to candidate query fields,
matching any operand not matched with a candidate query field to a default query field specified in the query parameters, and
adding the matched candidate queries and operands to the query plan as conditional parameters for the query,
wherein, in response to identifying implicit ordering criteria based on time, the computer system adapts ordering criteria based on the user's query to ensure results that have immediate time significance are placed ahead of less-immediate results; and
creating, based on the query plan, a database query, wherein the database query is in a database query language.
12. The system of claim 11 , further comprising adding a record count limit and ordering criteria to the query plan.
13. The system of claim 11 , wherein the system identifies any queryable field located between an interrogative or a lookup action entity and an entity corresponding to the database object as a subject field.
14. The system of claim 11 , wherein prior to determining if any of the candidate query fields are subject fields, the system preprocesses the query fields and operands.
15. The system of claim 14 , wherein preprocessing the operands and query fields comprises removing any redundant, trivial, and subsumed query fields and operands from the candidate query fields and operands.