Automated Patent Landscaping System Demonstrates Improved Performance

Strategic Insights -

Research conducted by Florida International University has presented an automated neural patent landscaping system that improves patent assessment and intellectual property context in a more efficient and effective manner. The system employs a minimal number of labeled examples, reducing the need for highly specialized knowledge and technical expertise. Funded by the National Science Foundation, this research contributes to the development of more accurate and comprehensive patent landscapes.

Key Takeaways:

  • The research highlights the importance of patent landscaping in assessing intellectual property contexts, particularly in high-tech areas.
  • The automated neural patent landscaping system demonstrated improved performance on difficult examples, achieving an F1 score of 0.69 on 'hard' examples, compared to 0.6 for previously reported systems.
  • The system also achieved significant improvements using much less training data, with an overall F1 score of 0.75 on as few as 24 examples.
  • The research employed a higher-quality training data generation procedure by merging the 'seed/anti-seed' approach with active learning to collect difficult labeled examples near the decision boundary.
  • A new dataset of labeled AI patents was created for training and testing, which is released for others to build upon under the Creative Commons NC-BY 4.0 license.
  • The paper was published in the journal Artificial Intelligence and Law and is available on the Springer website.

Statistics:

  • The automated neural patent landscaping system demonstrated a 0.09 improvement in F1 score on difficult examples compared to previous systems.
  • The system achieved significant improvements using much less training data, with a 24-example threshold for high accuracy.
  • The research utilized a higher-quality training data generation procedure, which resulted in a more comprehensive dataset of labeled AI patents.

Sources:

  • Artificial Intelligence and Law, 2025
  • Automated Neural Patent Landscaping In the Small Data Regime Using Citations and Cpc Codes
  • VerticalNews story, 2025
  • Florida International University Press Release
  • Published in Robotics & Machine Learning, October 27, 2025; p 472.