Artificial intelligence (AI) is rapidly changing the world around us. From self-driving cars to facial recognition software, AI is already having a major impact on our lives. But what about the use of AI in patent law?

Patent Validity/Invalidity Searches

Patent validity/invalidity searches are a critical part of the patent process. These searches are used to determine whether a patent is valid or invalid. This information is then often used to invalidate an existing patent.

Traditional patent validity/invalidity searches are conducted by human experts. However, AI is now being used to automate some of the tasks involved in these searches. While AI has the potential to revolutionize the patent search process, it also has some limitations.

Transformative Capabilities, Critical Limitations

In this blog post, we will discuss some of the key limitations of AI models for patent validity/invalidity searches. We will also discuss some of the ways that these limitations can be addressed.

Limitations of AI Models in Patent Validity/Invalidity Searches

  1. Bias in Training Data

AI models are only as good as the data they are trained on. If the training data is biased, then the AI model will also be biased. This can lead to inaccurate results.

For example, if an AI model is trained on a dataset of patents that is not representative of all jurisdictions or technical domains, then the model may be biased towards certain jurisdictions or technical domains. This could lead to the model making inaccurate predictions about the validity of patents in other jurisdictions or technical domains.

  1. Complexity of Patent Language

Patent language is often very complex and difficult to understand. AI models may struggle to interpret this language correctly. This can lead to inaccurate results.

  1. Fluctuating Legal Frameworks

Patent law is constantly evolving. AI models that are not updated regularly may not be able to account for these changes. This can lead to inaccurate results.

  1. Limited Understanding of Patent Drawings

Many patents include drawings or diagrams. AI models may not be able to understand these drawings correctly. This can lead to inaccurate results.

  1. Challenges in Evaluating Novelty and Inventive Step

AI models may struggle to evaluate the novelty and inventive step of a patent. These are two of the key criteria that are used to determine whether a patent is valid.

How to Address These Limitations

There are a number of ways to address the limitations of AI models in patent validity/invalidity searches.

  1. Training Data

AI models are heavily reliant on the quality and breadth of their training data. If the data is skewed, incomplete, or biased, the AI’s output will likely reflect those shortcomings. This is particularly crucial in patent analysis, where a diverse range of technologies, legal precedents, and global patent systems need to be considered.

Solutions:

  • Diversity: Training datasets should encompass a wide array of patent documents from various jurisdictions and technical fields to minimize bias and ensure the AI model can generalize effectively.
  • Volume: Large datasets, potentially exceeding a million patent documents, including both historical and contemporary data, are crucial to capture the evolving nature of technology and legal standards.
  • Refinement: Continuous refinement of the AI model by incorporating feedback from patent examiners, legal experts, and technical specialists helps in reducing bias and enhancing accuracy.
  1. Patent Language

Patent language is known for its complexity and nuanced phrasing, making it challenging for AI models to interpret accurately.

Solutions:

  • Advanced NLP: Employing sophisticated natural language processing (NLP) techniques like dependency parsing and semantic role labelling can help AI models better dissect the intricacies of patent claims.
  • Hybrid Models with Explainable AI (XAI): Integrating XAI into AI models provides transparency into the model’s reasoning process, allowing analysts to understand how the AI arrived at its conclusions and fostering greater trust in the results.
  • Contextual Enhancement: Utilizing specialized ontologies and knowledge graphs can provide the AI model with the necessary context to understand the technical and legal nuances embedded within patent claims.
  1. Legal Framework

Patent laws and precedents are constantly evolving, posing a challenge for AI models trained on static datasets.

Solutions:

  • Dynamic Legal Knowledge Graphs: Incorporating legal knowledge graphs that map evolving case law and statutory changes can help AI models stay current.
  • Real-time Updates: Integrating real-time alerts for significant legal updates ensures the AI model is aware of the latest changes in patent law.
  • Human Oversight: Human oversight remains crucial to ensure that AI-generated findings align with the latest legal standards and interpretations.
  • Litigation Data: Incorporating insights from patent litigation outcomes, such as court interpretations of claim scope and validity rulings, can enhance the AI’s ability to predict potential challenges.
  1. Patent Drawings

Many patents rely on intricate diagrams and technical drawings, which traditional AI models often struggle to interpret.

Solutions:

  • Computer Vision: Integrating computer vision techniques into AI models can enable them to analyse and extract insights from visual elements in patent drawings.
  • Annotated Training Data: Training AI systems on annotated patent diagrams, such as protein sequences or chemical structures, can enhance their ability to interpret visual data.
  • Human Validation: Human review can complement AI analysis by validating the model’s interpretations of technical illustrations.
  1. Novelty and Inventive Step

Assessing novelty and inventive step, which vary by jurisdiction, is a significant challenge for AI models.

Solutions:

  • Preliminary Assessments: AI can be used to conduct preliminary assessments of novelty and inventive step, which are then reviewed by humans.
  • Transparent Decision-Making: Enhancing transparency in AI decision-making processes using XAI allows human evaluators to understand and refine the AI’s reasoning.

Conclusion

AI has the potential to revolutionize the patent search process. However, it is important to be aware of the limitations of AI models. By addressing these limitations, we can ensure that AI is used in a responsible and ethical manner.

The future of patent searching lies in a hybrid approach that combines the strengths of AI with the expertise of human patent professionals. This approach will allow us to harness the full potential of AI while still ensuring the accuracy and reliability of patent searches.

About the author

Nouiere Järvinen is the Co-Founder and Chief Product Officer of Litigence, where she leads the development of AI-driven solutions that simplify patent processes and legal complexities. With a strong foundation in software engineering and a passion for technology’s role in intellectual property, she specializes in leveraging Generative AI to advance patent analysis and strategy. Her expertise spans invalidity searches, infringement analysis, novelty assessments, and litigation support. Her technical expertise, coupled with her commitment to AI-powered IP solutions, enables her to empower corporations to secure and maximize the value of their innovations. She is dedicated to making patent law more accessible, efficient, and impactful for global innovators.