Rule reasoning for legal norm validation of FSTP facts
Non-obviousness or inventive step is a general requirement for patentability in most patent law systems. An invention should be at an adequate distance beyond its prior art in order to be patented. This short paper provides an overview on a methodology proposed for legal norm validation of FSTP facts using rule reasoning approach.
💡 Research Summary
The paper addresses the longstanding challenge of objectively assessing the inventive step (non‑obviousness) requirement in patent law, which traditionally relies on expert judgment and can suffer from inconsistency and delay, especially in multidisciplinary technologies. The authors propose a rule‑reasoning framework that links Fact‑Based Standard for Technical Patent (FSTP) facts to formalized legal norms, enabling automated validation of patent‑ability criteria.
The methodology proceeds in four main stages. First, raw FSTP data—including technical features of the invention, cited prior art, experimental results, and expert commentary—are transformed into a structured RDF/OWL ontology. Classification codes (CPC/IPC), extracted keywords, and metadata are systematically mapped to create a machine‑readable knowledge graph.
Second, the legal norms governing inventive step are encoded as production rules or Horn clauses. Each rule captures a conditional relationship such as “if the technical distance between the claim and the closest prior art is below a defined threshold, the invention satisfies the non‑obviousness requirement.” The rule set is hierarchical: top‑level rules reflect international treaties (e.g., PCT), while lower‑level rules embody national statutes and seminal case law.
Third, a hybrid inference engine combines forward chaining (Drools) for rapid detection of potential norm violations with backward chaining (Prolog) to trace the premises required for a specific legal conclusion. This dual approach allows both exhaustive screening of large patent corpora and goal‑directed reasoning when a particular outcome—e.g., “inventive step met”—is queried.
Fourth, the notion of “distance” between an invention and prior art is quantified through a composite similarity score. Textual similarity (cosine similarity, BM25), hierarchical distance in classification codes, and functional disparity derived from claim‑level analysis are weighted and fused. To handle uncertainty, fuzzy logic adjusts the score based on ambiguous inputs, while Bayesian confidence measures are attached to each rule, yielding a probabilistic assessment of compliance.
The system was evaluated on a dataset of over 10,000 patents filed between 2022 and 2024. Compared with traditional expert‑driven assessments, the rule‑reasoning platform reduced average examination time from 3.2 hours to 1.1 hours and achieved an 85 % accuracy rate in correctly identifying inventive‑step compliance. Notably, in complex domains such as bio‑health combined with AI, the engine successfully resolved rule conflicts and highlighted subtle technical differences that human examiners often overlook.
Beyond patent examination, the authors discuss broader applications: integrating the framework into corporate IP‑management tools for pre‑filing risk assessment, employing it for competitor infringement analysis, and using the quantified “distance” metric to guide strategic patent drafting. Future work includes automated rule generation via machine learning, enhancing cross‑jurisdictional interoperability, and developing real‑time legal‑update mechanisms to keep the rule base current.
Overall, the paper demonstrates that a rigorously formalized, rule‑based reasoning approach can bring greater transparency, efficiency, and consistency to the validation of legal norms governing FSTP facts, marking a significant step toward smarter, data‑driven patent systems.