Search Planning Based on the Type of Storage: Searching for Composite Keys in LLM-Based Corporate Knowledge Systems
American Impact Review · 2026
DOI: 10.66308/air.e2026072 ↗Kurz gefasst
Fachartikel in der Zeitschrift American Impact Review, 2026. Verfasst von Mykola Khoroshevskyi. Die Arbeit nennt 10 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
The article describes the planning of search in corporate knowledge systems using large language models, taking into account the structure of the data warehouse. Special attention is paid to the search for objects that are identified by composite keys. To do this, you need to take into account the data schema, the relationships between objects, the available indexes, and the order in which sources are accessed. The article discusses various approaches to search: lexical, vector, hybrid, federated, and multi-stage. The architecture of LLM-guided search and the algorithm for processing complex queries are also presented. It is shown that large language models can be used to interpret queries, select search actions, and refine subsequent operations. However, information about the data structure must come from the metadata of the repository. To evaluate the proposed solutions, BEAVER and HERB enterprise suites are being considered, which include verification of work with structured and heterogeneous data sources.
Zitieren
APA
Khoroshevskyi, M. (2026). Search Planning Based on the Type of Storage: Searching for Composite Keys in LLM-Based Corporate Knowledge Systems. American Impact Review. https://doi.org/10.66308/air.e2026072
Vancouver
Khoroshevskyi M. Search Planning Based on the Type of Storage: Searching for Composite Keys in LLM-Based Corporate Knowledge Systems. American Impact Review. 2026. doi:10.66308/air.e2026072
BibTeX
@article{Khoroshevskyi2026,
author = {Khoroshevskyi, Mykola},
title = {Search Planning Based on the Type of Storage: Searching for Composite Keys in LLM-Based Corporate Knowledge Systems},
journal = {American Impact Review},
year = {2026},
doi = {10.66308/air.e2026072}
}
EUVIMED