A Purpose-Aware Semantic Reasoning Model for Patent Infringement Detection in the DIKWP Network

Patent infringement detection requires coordinated interpretation of technical claims, legal standards, and contextual evidence. This study proposes a semantic AI framework for patent infringement detection grounded in the DIKWP network and artificial consciousness theory. The DIKWP network organizes the analytical modules as interacting semantic spaces rather than as a strictly layered pipeline. This design supports iterative semantic interpretation, knowledge integration, and purpose-oriented reasoning. The framework integrates document ingestion, semantic information extraction, ontology-based knowledge representation, rule-guided inference, and decision support. The system processes patent claims, product descriptions, and prior-art documents with patent-oriented NLP. Named entity recognition and subject–action–object parsing convert unstructured text into structured semantic representations. Legal and technical ontologies support claim-element interpretation. Knowledge graphs, semantic pattern matching, and inference rules then align claim elements with product features and identify potential infringement risks. A prototype implementation demonstrates end-to-end processing from raw text to infringement-oriented assessment. The evaluation was conducted in two layers. First, a controlled synthetic patent–product corpus was used to isolate claim-element reasoning, rule-guided inference, and purpose-conditioned operating modes. Second, a real-world pilot corpus was constructed from publicly available patent claims and real product technical descriptions, including manufacturer manuals, technical datasheets, official product webpages, installation guides, and technical brochures. The controlled-corpus results show that the DIKWP network improves over keyword-matching and ontology-only baselines by integrating semantic coverage, claim-level legal reasoning, and explainable output. The real-world pilot provides a preliminary external-validity check of whether the framework can preserve element-level reasoning under realistic drafting styles, domain terminology, incomplete product evidence, and borderline claim-to-product correspondences. These findings provide preliminary evidence of feasibility and analytical value, rather than a final benchmark of litigation-level performance.