Abstract
ABSTRACT
Systems, methods, and computer program products are provided for processing documents. The system includes at least one processor configured to segment a document into a plurality of segments, construct a hierarchical data structure comprising a plurality of nodes representing the plurality of segments, each node of the plurality of nodes comprising hierarchical metadata, in response to determining at least one first node of the plurality of nodes exceeds a token limit, split the at least one first node into a first subset of nodes based on a semantic analysis of text represented by the at least one first node, merge at least two nodes of the plurality of nodes into a single node based on the hierarchical metadata for the at least two nodes, the at least two nodes sharing a direct parent node, and input at least a subset of nodes of the plurality of nodes into a large language model (LLM) with a corresponding prompt.
FIG. 1
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
MAHTO, KOMAL, "SYSTEM, METHOD, AND COMPUTER PROGRAM PRODUCT FOR PROCESSING DOCUMENTS INTO HIERARCHICAL DATA STRUCTURES FOR RETRIEVAL-AUGMENTED GENERATION", Technical Disclosure Commons, (July 06, 2026)
https://www.tdcommons.org/dpubs_series/10813