Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

SHRAG: AFrameworkfor Combining Human-Inspired Search with RAG

Hyunseok Ryu, Wonjune Shin, Hyun Park

TL;DR

This work tackles the practical inefficiencies of Retrieval-Augmented Generation by presenting SHRAG, a plug-and-play framework that uses an LLM as a Query Strategist to convert natural language queries into structured, Boolean-like search queries. By coupling multilingual keyword extraction, OR-based strategic query expansion, cross-lingual embedding, and selective answer generation, SHRAG reduces latency and cost while preserving retrieval reliability. The authors demonstrate strong results on the ScienceON AI Challenge, including first-place performance, and validate generalization with the MIRACL dataset, highlighting robust cross-lingual evidence gathering. Overall, SHRAG advocates a pragmatic shift in RAG design that leverages existing search infrastructure to deliver direct, evidence-backed multilingual responses without heavy fine-tuning.

Abstract

Retrieval-Augmented Generation (RAG) is gaining recognition as one of the key technological axes for next generation information retrieval, owing to its ability to mitigate the hallucination phenomenon in Large Language Models (LLMs)and effectively incorporate up-to-date information. However, specialized expertise is necessary to construct ahigh-quality retrieval system independently; moreover, RAGdemonstratesrelativelyslowerprocessing speeds compared to conventional pure retrieval systems because it involves both retrieval and generation stages. Accordingly, this study proposes SHRAG, a novel framework designed to facilitate the seamless integration of Information Retrieval and RAG while simultaneously securing precise retrieval performance. SHRAG utilizes a Large Language Model as a Query Strategist to automatically transform unstructured natural language queries into logically structured search queries, subsequently performing Boolean retrieval to emulate the search process of an expert human searcher. Furthermore, it incorporates multilingual query expansion and a multilingual embedding model, enabling it to perform efficient cross-lingual question answering within the multilingual dataset environment of the ScienceON Challenge. Experimental results demonstrate that the proposed method, combining logical retrieval capabilities and generative reasoning, can significantly enhance the accuracy and reliability of RAG systems. Furthermore, SHRAG movesbeyondconventionaldocument-centric retrieval methods, presenting the potential for a new search paradigm capable of providing direct and reliable responses to queries.

SHRAG: AFrameworkfor Combining Human-Inspired Search with RAG

TL;DR

This work tackles the practical inefficiencies of Retrieval-Augmented Generation by presenting SHRAG, a plug-and-play framework that uses an LLM as a Query Strategist to convert natural language queries into structured, Boolean-like search queries. By coupling multilingual keyword extraction, OR-based strategic query expansion, cross-lingual embedding, and selective answer generation, SHRAG reduces latency and cost while preserving retrieval reliability. The authors demonstrate strong results on the ScienceON AI Challenge, including first-place performance, and validate generalization with the MIRACL dataset, highlighting robust cross-lingual evidence gathering. Overall, SHRAG advocates a pragmatic shift in RAG design that leverages existing search infrastructure to deliver direct, evidence-backed multilingual responses without heavy fine-tuning.

Abstract

Retrieval-Augmented Generation (RAG) is gaining recognition as one of the key technological axes for next generation information retrieval, owing to its ability to mitigate the hallucination phenomenon in Large Language Models (LLMs)and effectively incorporate up-to-date information. However, specialized expertise is necessary to construct ahigh-quality retrieval system independently; moreover, RAGdemonstratesrelativelyslowerprocessing speeds compared to conventional pure retrieval systems because it involves both retrieval and generation stages. Accordingly, this study proposes SHRAG, a novel framework designed to facilitate the seamless integration of Information Retrieval and RAG while simultaneously securing precise retrieval performance. SHRAG utilizes a Large Language Model as a Query Strategist to automatically transform unstructured natural language queries into logically structured search queries, subsequently performing Boolean retrieval to emulate the search process of an expert human searcher. Furthermore, it incorporates multilingual query expansion and a multilingual embedding model, enabling it to perform efficient cross-lingual question answering within the multilingual dataset environment of the ScienceON Challenge. Experimental results demonstrate that the proposed method, combining logical retrieval capabilities and generative reasoning, can significantly enhance the accuracy and reliability of RAG systems. Furthermore, SHRAG movesbeyondconventionaldocument-centric retrieval methods, presenting the potential for a new search paradigm capable of providing direct and reliable responses to queries.

Paper Structure

This paper contains 34 sections, 7 equations, 4 figures, 1 table, 1 algorithm.

Table of Contents

  1. Introduction
  2. Preliminaries
  3. Retrieval-Augmented Generation (RAG)
  4. Dense and Sparse Retrieval
  5. Multilingual Embedding
  6. Single-Hop and Multi-Hop
  7. Boolean Retrieval
  8. Methodology: Search like human with RAG (SHRAG)
  9. Input and Output
  10. Step 1: Multilingual Keyword Extraction
  11. Step 2: Strategic Query Generation
  12. Step 3: Document Retrieval and Deduplication
  13. Step 4: Multilingual Re-ranking
  14. Step 5: Structured Answer Generation
  15. Multilingual Keyword Extraction
  16. ...and 19 more sections

Figures (4)

  • Figure 1: An illustration explaining the system's overall structure and components.
  • Figure 2: An illustration explaining Multilingual Keyword Extraction and Search Query Set generated by Strategic Search Query Generation.
  • Figure 3: Left: The proportion of relevant documents that were successfully searched, categorized by the count of AND operators. Right: Average number of searched documents (over all queries) categorized by the count of AND operators.
  • Figure 4: Overall system architecture with Query Decomposer for multi-hop query processing