Evidence retrieval for long document question answering using large language models
AssigneeAdobe Inc.
InventorsInderjeet NAIR, Shwetha S, Apoorv SAXENA, Koustava GOSWAMI
Life of this patent
Estimating the term…
Abstract
Embodiments are disclosed for long document question answering using large language models. The method may include receiving a question for a document. A representation of the document may then be obtained. A large language model (LLM) is used to identify one or more sections of the document that are relevant to the question using the representation of the document. A document question answering model determines an answer to the question using the one or more sections of the document.
Claims20
A method comprising:
- receiving a question for a document;
- obtaining a representation of the document;
- identifying, using a large language model (LLM), one or more sections of the document that are relevant to the question using the representation of the document;
- and
- determining, by a document question answering model, an answer to the question using the one or more sections of the document.
A non-transitory computer-readable medium storing executable instructions, which when executed by a processing device, cause the processing device to perform operations comprising:
- receiving a question for a document;
- obtaining a representation of the document;
- identifying, using a large language model (LLM), one or more sections of the document that are relevant to the question using the representation of the document;
- and
- determining, by a document question answering model, an answer to the question using the one or more sections of the document.
A system comprising:
- a memory component;
- and
- a processing device coupled to the memory component, the processing device to perform operations comprising:
- receiving a question for a document;
- generating, by a summary generator, a representation of the document;
- identifying, using a coarse-grain large language model (LLM), one or more relevant sections of the document to the question using the representation of the document;
- identifying, using a fine-grain LLM, one or more relevant paragraphs of the document that are relevant to the question based on the relevant sections;
- and
- determining, by a document question answering model, an answer to the question using document text corresponding to the one or more relevant paragraphs.
Dependencies are inferred from the claim text.