FlowQuery: AI-Assisted Visual Boolean Query Construction for Iterative Patent Search Development

Patent search requires precise and often complex Boolean query construction to retrieve relevant prior art across vast, linguistically diverse datasets. This process remains cognitively demanding and error-prone, with traditional text-based interfaces offering little support for iterative refinement or semantic exploration. We present a novel system that reimagines patent search as an interactive, visual process combining symbolic precision with semantic guidance. Our system features a graphical Boolean query editor, AI-assisted refinement workflows, vector-based semantic clustering, and drag-and-drop interactivity from results to query space. The visual interface enhances transparency by eliminating ambiguities in operator precedence and supporting nested, foldable logic containers. Crucially, our system reduces friction in the discovery and refinement loop: users can drag significant terms, date ranges, and even semantic clusters directly from search results into structured queries. These features promote a fluid interaction model, where exploratory actions yield immediate, actionable refinements. The system supports integration with varied backend search engines via minimal adaptation, ensuring wide applicability. By bridging the gap between exploratory search and legal-grade precision, our approach significantly improves usability, transparency, and effectiveness in patent information retrieval. This paper outlines the system's architecture, interaction model, and potential to reshape workflows for legal and technical users alike.