Foresight analysis relies on structured knowledge retrieval and systematic scenario-building to anticipate future developments in complex domains. Traditional methods require extensive expert involvement in ontology construction, data classification, and uncertainty resolution, often making the process resource-intensive and time-consuming. This study introduces a semantic role-based querying approach utilizing Large Language Models (LLMs) to enhance adaptive subject area structuring for foresight applications. By embedding explicit semantic roles (e.g., agent, cause, effect, instrument) directly into query prompts, this approach reverses traditional Semantic Role Labeling (SRL), ensuring context-aware, structured knowledge retrieval. Furthermore, this approach was validated by constructing a novel cryptocurrency market classifier, demonstrating its applicability in generating previously non-existent structured taxonomies with explicit semantic dependencies. The study evaluates the effectiveness of this method through controlled experimental analysis, comparing 5-point and 10-point structured queries across multiple runs. Results demonstrate that semantic role-based querying significantly improves response consistency, hierarchical organization, and relevance in foresight-driven knowledge retrieval. While higher requested points per query provide condensed, high-level categorization, lower requested points (aggregated across runs) capture greater diversity and institutional specificity. Additionally, the second experiment illustrates how reverse SRL can be leveraged to extract functionally connected classifications, ensuring precise differentiation between causal agents, technological enablers, regulatory constraints, and economic influences within a complex domain. These findings highlight the stability, adaptability, and efficiency of LLM-driven foresight analysis, reducing manual effort while maintaining expert oversight. The proposed approach enhances predictive accuracy on early foresight stages under decision-support frameworks and strategic scenario modeling, marking a significant step toward AI-assisted foresight methodologies.