Method and System for Optimizing Use of Retrieval Augmented Generation Pipelines in Generative Artificial Intelligence Applications

A computer-implemented system for cross-domain structural pattern matching in a retrieval augmented generation pipeline, comprising:

• a function identification module configured to analyze a plurality of documents across a plurality of domains using domain-specific analyzers and identify functional components within each domain, wherein each functional component comprises a function representing a domain-specific task, at least one event representing an outcome of said function, and at least one way representing an implementation method for said function;
• a graph database encoding module configured to encode structural relationships between said identified functional components in a graph database using structural labels that capture functional abstractions independent of domain-specific terminology, wherein said encoding captures hierarchical dependencies between functions, events, and ways within and across said plurality of domains independent of domain-specific surface vocabulary;
• and
• a structural retrieval module configured to compute structural similarity between a target query and functional components encoded in said graph database, and retrieve one or more functional components whose structural similarity to said target query exceeds a predetermined threshold for augmenting generation of a response by a large language model.

A system for performing cross-domain structural pattern matching in a retrieval augmented generation, comprising:

• a processor;
• and
• a non-transitory computer-readable storage medium having stored thereon computer-executable instructions that, when executed by the processor, causes the processor to perform operations comprising:
• analyzing a plurality of documents across a plurality of domains using domain-specific analyzers to identify one or more functional components within each domain, each functional component comprising a function representing a domain-specific task or process, at least one event representing an outcome of said function, and at least one way representing an implementation method for said function;
• encoding structural relationships between said functional components in a graph database independent of domain-specific surface vocabulary, wherein said encoding captures hierarchical dependencies between functions, events, and ways within and across said plurality of domains;
• computing structural similarity between a target query and said encoded functional components based on said one or more structural relationships;
• and
• retrieving one or more functionally similar functional components having a structural similarity that exceeds a predetermined threshold for augmenting generation of a response to said target query by a large language model.