Data and Methods for Identifying Artificial Intelligence Patents
New data analysis techniques developed for identifying AI-related patents.
1Importance of AI Patents
Rapid advancements in AI technologies have led to a significant increase in AI patents. Accurate identification of these patents has become crucial for understanding tech trends and gaining competitive advantage. Researchers found traditional patent classification methods inadequate for AI patents. This gap highlighted the need for new analytical techniques in this emerging field.
2Limitations of Current Methods
Traditional patent classification systems fail to capture the diversity of AI technologies. Existing methods struggle to distinguish between subfields like machine learning and deep learning. The complexity of technical jargon in patent texts further complicates accurate classification. These limitations hinder effective tracking of AI patent activity across industries.
3Novel Data Analysis Approach
Researchers developed a new data analysis method specifically for identifying AI patents. The technique more precisely detects keywords and technical terms in patent texts. Using machine learning algorithms, it quantifies the AI content level of patents. The method employs domain-specific language models for different AI subfields.
4Dataset and Scope
The study analyzed 1.2 million patents granted between 2010-2025. Data was collected from patent offices in the US, Europe, and Asia. Researchers examined not only patent texts but also citation networks and inventor information. This comprehensive dataset enhanced the method's accuracy in identifying true AI patents.
5Algorithm Performance
The new algorithm demonstrated 35% higher accuracy than traditional methods. For deep learning patents specifically, classification success reached 92%. The system effectively recognized different languages and technical terminologies. Performance was validated across multiple independent test datasets.
6Industry Applications
This method enables tech companies to monitor competitors' AI investments. Patent attorneys can better evaluate clients' AI portfolios. Investors gain earlier detection of emerging AI trends. Policymakers can create more accurate AI innovation maps for strategic planning purposes.
7Future Developments
Researchers suggest the method could be adapted to other technology domains. Similar analyses could be applied to blockchain and quantum computing patents. Making the tool open-source will encourage wider adoption. Future work will focus on real-time patent monitoring systems.
8Mapping AI Innovation
This study represents a significant step in understanding global AI innovation ecosystems. The developed methods provide valuable insights for technology transfer and R&D strategies. Accurate classification of AI patents will help shape national innovation policies. Findings were published in the November 2026 issue of Research Policy.
Original article: Research Policy — Yapay Zeka Patentlerini Tanımlamak için Veri ve Yöntemler