Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

FlexiDataGen: An Adaptive LLM Framework for Dynamic Semantic Dataset Generation in Sensitive Domains

Hamed Jelodar, Samita Bai, Roozbeh Razavi-Far, Ali A. Ghorbani

TL;DR

This work tackles data scarcity and privacy constraints in high-stakes domains by introducing FlexiDataGen, an adaptive LLM-based framework for dynamic semantic dataset generation. The approach combines five phases—syntactic-semantic analysis, retrieval-augmented generation ($RAG$), semantic scenario definition, dynamic prompt construction, and iterative paraphrasing with semantic validation—to autonomously produce semantically rich domain prompts. Key contributions include a modular pipeline that integrates $RAG$ with domain grounding, semantic scenario instantiation, and paraphrase validation to maintain fidelity while enhancing diversity, demonstrated in medical use cases. The results indicate robust dataset generation with high uniqueness and low noise, enabling scalable prompt tuning, data augmentation, and evaluation in sensitive fields while preserving privacy and domain fidelity.

Abstract

Dataset availability and quality remain critical challenges in machine learning, especially in domains where data are scarce, expensive to acquire, or constrained by privacy regulations. Fields such as healthcare, biomedical research, and cybersecurity frequently encounter high data acquisition costs, limited access to annotated data, and the rarity or sensitivity of key events. These issues-collectively referred to as the dataset challenge-hinder the development of accurate and generalizable machine learning models in such high-stakes domains. To address this, we introduce FlexiDataGen, an adaptive large language model (LLM) framework designed for dynamic semantic dataset generation in sensitive domains. FlexiDataGen autonomously synthesizes rich, semantically coherent, and linguistically diverse datasets tailored to specialized fields. The framework integrates four core components: (1) syntactic-semantic analysis, (2) retrieval-augmented generation, (3) dynamic element injection, and (4) iterative paraphrasing with semantic validation. Together, these components ensure the generation of high-quality, domain-relevant data. Experimental results show that FlexiDataGen effectively alleviates data shortages and annotation bottlenecks, enabling scalable and accurate machine learning model development.

FlexiDataGen: An Adaptive LLM Framework for Dynamic Semantic Dataset Generation in Sensitive Domains

TL;DR

This work tackles data scarcity and privacy constraints in high-stakes domains by introducing FlexiDataGen, an adaptive LLM-based framework for dynamic semantic dataset generation. The approach combines five phases—syntactic-semantic analysis, retrieval-augmented generation (), semantic scenario definition, dynamic prompt construction, and iterative paraphrasing with semantic validation—to autonomously produce semantically rich domain prompts. Key contributions include a modular pipeline that integrates with domain grounding, semantic scenario instantiation, and paraphrase validation to maintain fidelity while enhancing diversity, demonstrated in medical use cases. The results indicate robust dataset generation with high uniqueness and low noise, enabling scalable prompt tuning, data augmentation, and evaluation in sensitive fields while preserving privacy and domain fidelity.

Abstract

Dataset availability and quality remain critical challenges in machine learning, especially in domains where data are scarce, expensive to acquire, or constrained by privacy regulations. Fields such as healthcare, biomedical research, and cybersecurity frequently encounter high data acquisition costs, limited access to annotated data, and the rarity or sensitivity of key events. These issues-collectively referred to as the dataset challenge-hinder the development of accurate and generalizable machine learning models in such high-stakes domains. To address this, we introduce FlexiDataGen, an adaptive large language model (LLM) framework designed for dynamic semantic dataset generation in sensitive domains. FlexiDataGen autonomously synthesizes rich, semantically coherent, and linguistically diverse datasets tailored to specialized fields. The framework integrates four core components: (1) syntactic-semantic analysis, (2) retrieval-augmented generation, (3) dynamic element injection, and (4) iterative paraphrasing with semantic validation. Together, these components ensure the generation of high-quality, domain-relevant data. Experimental results show that FlexiDataGen effectively alleviates data shortages and annotation bottlenecks, enabling scalable and accurate machine learning model development.
Paper Structure (14 sections, 8 equations, 4 figures, 2 tables, 2 algorithms)

This paper contains 14 sections, 8 equations, 4 figures, 2 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. FlexiDataGen Framework
  3. Phase 1: Input and Pattern Analysis
  4. Phase 2: Semantic Scenario Definition
  5. Phase 3: Dynamic Element Injection and Prompt Construction
  6. Phase 4: Iterative Paraphrasing and Semantic Similarity Validation
  7. Phase 5: Dataset Generation and Output
  8. Experiment
  9. Computing Infrastructure
  10. LLM Hyper-parameters
  11. Use Cases and Applications
  12. Model Performance in the Medical Use Case
  13. Discussion
  14. Conclusion

Figures (4)

  • Figure 1: Application scenarios for domain-adaptive data generation using FlexiDataGen
  • Figure 2: Conceptual Diagram of the Adaptive LLM Framework for Dynamic Semantic Dataset Generation.
  • Figure 3: The framework comprises two key components: the first focuses on constructing the prompt dataset, while the second is dedicated to generating the final sample dataset.
  • Figure 4: An Example of Semantic Prompt Tree for EHR (Electronic Health Records) Scenario Generation.