Back to results
US 12,761,979 B2Granted
Large language model and deterministic calculator systems and methods
AssigneeIntuit Inc.
InventorsNa Xu, Meng Chen, Conrad De Peuter
Life of this patent
Estimating the term…
Abstract
At least one large language model (LLM) may be instructed to request data from a user while being prevented from performing calculations using the data. A user-generated response to the request for data, including at least a portion of the data, may be received. The user-generated response may be translated into a machine-readable response in a format configured for processing by a calculation engine. The calculation engine may process the machine-readable response, thereby generating a calculation engine output.
Claims20
A method comprising:
- providing a predefined data model to at least one large language model (LLM), the predefined data model defining data required to perform a calculation for a user;
- generating, by at least one processor, an instruction configured to cause the at least one LLM to request data to fill in the predefined data model from the user while preventing the at least one LLM from performing calculations using the data;
- presenting, by the at least one processor, a user interface (UI) including output of the at least one LLM to the user, wherein the request for data by the at least one LLM is made through the UI;
- receiving, through the UI, a user-generated response to the request for data, the user-generated response including at least a portion of the data;
- translating the user-generated response into a machine-readable response in a format configured for processing by a calculation engine executed by the at least one processor;
- processing, by the calculation engine executed by the at least one processor, the machine-readable response, thereby generating a calculation engine output;
- and
- modifying, by the at least one processor, the UI to include an indication of the calculation engine output.
5 more elements
A method comprising:
- providing a predefined data model to at least one large language model (LLM), the predefined data model defining data required to perform a calculation for a user;
- generating, by at least one processor, an instruction configured to cause the at least one LLM to request data to fill in the predefined data model from the user while preventing the at least one LLM from performing calculations using the data;
- presenting, by the at least one processor, a user interface (UI) including output of the at least one LLM to the user;
- processing, by the LLM executed by the at least one processor, the instruction to thereby provide the request for data through the UI;
- receiving, through the UI, a user-generated response to the request for data, the user-generated response including at least a portion of the data;
- translating, by the at least one processor, the user-generated response into a machine-readable response in a format configured for processing by a calculation engine executed by the at least one processor;
- processing, by the calculation engine executed by the at least one processor, the machine-readable response, thereby generating a calculation engine output;
- and
- modifying, by the at least one processor, the UI to include an indication of the calculation engine output.
6 more elements
A system comprising:
- at least one processor;
- and
- at least one non-transitory computer-readable memory storing instructions that, when executed by the at least one processor, cause the at least one processor to perform processing comprising:
- providing a predefined data model to at least one large language model (LLM), the predefined data model defining data required to perform a calculation for a user;
- generating an instruction configured to cause at least LLM to request data to fill in the predefined data model from the user while preventing the at least one LLM from performing calculations using the data;
- presenting a user interface (UI) including output of the at least one LLM to the user, wherein the request for data by the at least one LLM is made through the UI;
- receiving a user-generated response to the request for data, the user-generated response including at least a portion of the data;
- translating the user-generated response into a machine-readable response in a format configured for processing by a calculation engine executed by the at least one processor;
- processing, by the calculation engine, the machine-readable response, thereby generating a calculation engine output;
- and
- modifying the UI to include an indication of the calculation engine output.
8 more elements
Dependencies are inferred from the claim text.