The Undeclared Patent Approach is designed to provide a more comprehensive view of the patent landscape by identifying patents that may be essential to specific technology standards but have not been formally declared.
This approach uses a classification algorithm, a machine learning–based method, to detect potentially undeclared standard-essential patents. The model is trained using both true positive examples (such as pooled patents) and true negative examples to build a broad representation of the relevant technology space. The process is iterative and incorporates input from subject matter experts, who validate patent essentiality and help refine the model’s accuracy over time.
By incorporating undeclared patents alongside declared standard-essential patents (SEPs) and non-SEPs, the approach enables a more complete exploration of the patent landscape. Users can further refine their analysis using tools such as the semantic essentiality score (SES), which helps filter and prioritise relevant results.
Overall, the Undeclared Patent Approach supports activities such as strategic decision-making, licensing analysis, and competitive benchmarking by offering a clearer and more detailed perspective on technology standards and associated patents.
The undeclared patent landscape is based on the following steps:
1. Define the Scope:
Begin by defining the scope of the technology area you are interested in. This involves identifying relevant keywords, concepts, and any themes that should be excluded. Providing a technology brief and examples of positive and negative data can help refine the search criteria.
2. Initial Classifier Building:
The initial phase involves uploading a pre-populated set of positive training patents and identifying similar patent families to source a negative training set. This helps the algorithm distinguish between closely related technologies.
3. Evaluation and Feedback:
The classifier is evaluated by internal experts to ensure it accurately captures relevant patents. Feedback from SEP holders is also incorporated to assess the quality and recall rates of the identified patents.
4. Refinement:
Based on the feedback, the classifier is refined to improve its precision and ensure it aligns with the defined scope of the project.
5. Search Execution:
Once the classifier is finalized, it is integrated into the IPlytics Platform. Users can then execute searches to identify undeclared patents within the specified technology landscape. The platform allows users to filter results by various criteria, such as the semantic essentiality score (SES), to further refine the search.
IPlytics has developed patent landscapes for undeclared patents for QI, video coding (AVC, HEVC, VVC) and Wifi (Wifi 4, Wifi 5, Wifi 6).