EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Leveraging NLP and Retrieval-Augmented Generation (RAG) Models for Automated Business Intelligence Query Resolution

Linda Aluso, Joy Onma Enyejo

International Journal of Scientific Research in Science, Engineering and Technology · 18.08.2024 · 11 (4) · 534-557

DOI: 10.32628/ijsrset242439 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Scientific Research in Science, Engineering and Technology, online erschienen am 18.08.2024. Verfasst von Linda Aluso und 1 weiteren Autor:innen. Die Arbeit wurde bisher 4-mal zitiert (Crossref). Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The rapid expansion of enterprise data ecosystems has intensified the demand for intelligent Business Intelligence (BI) systems capable of translating natural language queries into accurate analytical insights without requiring technical expertise. Traditional BI query interfaces rely heavily on structured query languages and predefined dashboards, creating bottlenecks in decision-making workflows and limiting accessibility for non-technical stakeholders. This paper proposes a novel framework titled Adaptive Contextual Retrieval-Augmented Intelligence Engine (ACRIE), which integrates advanced Natural Language Processing (NLP) with Retrieval-Augmented Generation (RAG) to automate business intelligence query resolution with improved semantic understanding, contextual grounding, and analytical accuracy. The proposed model combines transformer-based semantic parsing using Domain-Tuned BERT embeddings, hybrid vector–symbolic retrieval, and a dynamically optimized RAG pipeline that retrieves enterprise knowledge artifacts prior to response synthesis. Unlike conventional RAG implementations that rely solely on similarity search, ACRIE introduces a Context-Aware Query Reformulation Algorithm (CAQRA) and a Multi-Stage Evidence Ranking Mechanism (MSERM) to minimize hallucinations and enhance numerical reasoning during analytical query generation. Performance is evaluated against baseline systems including Text-to-SQL transformers, standard RAG architectures, and rule-based BI assistants across benchmark enterprise datasets. Experimental comparisons demonstrate measurable improvements in query resolution accuracy, latency reduction, and analytical consistency. The proposed framework achieves higher semantic alignment scores and improved retrieval precision while maintaining scalable inference performance suitable for real-time analytics environments. Graph-based evaluations and comparative performance visualizations illustrate gains in response correctness, contextual relevance, and decision-support reliability. The findings establish that integrating adaptive retrieval intelligence with domain-aware NLP significantly advances automated BI systems, enabling conversational analytics that bridge the gap between enterprise data warehouses and executive decision-making processes. The study contributes a scalable architecture and algorithmic innovation capable of redefining next-generation intelligent analytics platforms.

Zitieren

APA

Aluso, L., & Enyejo, J. O. (2024). Leveraging NLP and Retrieval-Augmented Generation (RAG) Models for Automated Business Intelligence Query Resolution. International Journal of Scientific Research in Science, Engineering and Technology, 11(4), 534-557. https://doi.org/10.32628/ijsrset242439

Vancouver

Aluso L, Enyejo JO. Leveraging NLP and Retrieval-Augmented Generation (RAG) Models for Automated Business Intelligence Query Resolution. International Journal of Scientific Research in Science, Engineering and Technology. 2024;11(4):534-557. doi:10.32628/ijsrset242439

BibTeX
@article{Aluso2024,
  author = {Aluso, Linda and Enyejo, Joy Onma},
  title = {Leveraging NLP and Retrieval-Augmented Generation (RAG) Models for Automated Business Intelligence Query Resolution},
  journal = {International Journal of Scientific Research in Science, Engineering and Technology},
  year = {2024},
  volume = {11},
  number = {4},
  pages = {534--557},
  doi = {10.32628/ijsrset242439}
}
Kostenlos starten →