Device and method for artificial intelligence reasoning based on language model
AssigneeElectronics and Telecommunications Research Institute
InventorYoung-Ae Seo
Life of this patent
Estimating the term…
Abstract
Provided are a device and method for artificial intelligence (AI) reasoning based on a language model. The method for AI reasoning based on a language model includes generating a question core summary by inputting a natural language question to a language model, decomposing the natural language question into sub-questions using a reasoning model on the basis of the question core summary and the natural language question, obtaining an answer to the natural language question by solving the sub-questions and synthesizing solutions of the sub-questions.
Claims12
A computer-implemented method for artificial intelligence (AI) reasoning based on a language model, the method comprising:
- inputting, by at least one processor, a natural language question to a pretrained language model to generate a question core summary, wherein the question core summary is a semi-code-style summary including variables corresponding to core information extracted from the natural language question, and wherein the semi-code-style summary represents information explicitly given in the natural language question separately from a reasoning process for answering the natural language question;
- decomposing, by the at least one processor, the natural language question into one or more sub-questions using a reasoning model by inputting both the semi-code-style summary and the natural language question to the reasoning model, wherein each of the one or more sub-questions corresponds to one reasoning operation;
- and
- solving, by the at least one processor, the one or more sub-questions and synthesizing solutions of the one or more sub-questions to obtain an answer to the natural language question.
A device for artificial intelligence (AI) reasoning based on a language model, the device comprising:
- a memory configured to store computer-readable instructions;
- and
- at least one processor configured to execute the instructions,
- wherein, when the instructions are executed, the at least one processor:
- inputs a natural language question to a pretrained language model to generate a question core summary, wherein the question core summary is a semi-code-style summary including variables corresponding to core information extracted from the natural language question, and wherein the semi-code-style summary represents information explicitly given in the natural language question separately from a reasoning process for answering the natural language question;
- decomposes the natural language question into one or more sub-questions using a reasoning model by inputting both the semi-code-style summary and the natural language question to the reasoning model, wherein each of the one or more sub-questions corresponds to one reasoning operation;
- and
- solves the one or more sub-questions, and obtains an answer to the natural language question by synthesizing solutions of the one or more sub-questions.
Dependencies are inferred from the claim text.