Abstract: A technique for interacting with a generative language model includes identifying one or more key sentences in an input document using a key sentence classifier model and/or an entity extraction model. Each key sentence summarizes a part of information conveyed by the input document. The technique further includes generating a compressed document that selectively includes the one or more key sentences. The technique then generates a prompt that includes the compressed document instead of the input document and submits the prompt to the language model. The technique reduces consumption of resources and increases performance by reducing the size of the prompt. A training system produces the key sentence classifier model by first training a pair-comparing model based on a relatively small amount of human-labeled data, and then leveraging the pair-comparing model to produce a synthetic data set on which the key sentence classifier model is trained.

Abstract: The disclosure herein describes generating input key-standard key mappings for a form. A set of input key-value pairs are received, and a subset of candidate form types are determined from a set of form types using the input key-value pairs. A set of standard keys associated with the determined subset of candidate form types are obtained. A set of input key-standard key pairs are generated using the set of input key-value pairs and the obtained set of standard keys and the set of input key-standard key pairs are narrowed using a narrowing rule. Ranking scores for each input key-standard key pair of the narrowed set of input key-standard key pairs are generated. Each input key of the set of input key-vale pairs is mapped to a standard key of the set of standard keys using at least the generated ranking scores of the narrowed set of input key-standard key pairs.

Abstract: The disclosure herein describes generating input key-standard key mappings for a form. A set of input key-value pairs are received, and a subset of candidate form types are determined from a set of form types using the input key-value pairs. A set of standard keys associated with the determined subset of candidate form types are obtained. A set of input key-standard key pairs are generated using the set of input key-value pairs and the obtained set of standard keys and the set of input key-standard key pairs are narrowed using a narrowing rule. Ranking scores for each input key-standard key pair of the narrowed set of input key-standard key pairs are generated. Each input key of the set of input key-vale pairs is mapped to a standard key of the set of standard keys using at least the generated ranking scores of the narrowed set of input key-standard key pairs.