Patent Claim Recommender for Technical Standards Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing IR systems fail to efficiently map patent claims to relevant sections of technical standards, particularly in large and evolving standards collections like ETSI, leading to high costs for patent owners in identifying infringers and monetizing their intellectual property.
Innovation Solution
A machine learning-based recommender system that uses feature extraction and dimensionality reduction techniques to generate feature vectors from patent claims and technical standards documents, followed by a binary classifier to rank-order relevance, utilizing ground truth datasets and user feedback for improved accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If traditional information retrieval systems are used to search for relevant technical standards documents, then the search process becomes simpler and faster, but the precision and relevance of mapping patent claims to specific standard sections deteriorates
Solution Approach 1:
The patent replaces traditional mechanical information retrieval systems with a machine learning-based recommender system. The system uses supervised learning algorithms trained on ground truth datasets to automatically map patent claims to relevant technical standard sections, substituting manual or simple keyword-based search mechanisms with intelligent computational models that achieve both speed and precision.
Solution Approach 2:
The patent transforms the search process by changing parameters from simple keyword matching to multi-dimensional feature analysis. The recommender system considers multiple features including claim text, standard document content, technical domains, and relevance scores to generate precise mappings, thereby improving measurement precision while maintaining search efficiency.
2Measurement precision
If manual methods are used to identify relevant sections of technical standards, then mapping precision improves, but productivity and time consumption deteriorate
Solution Approach 1:
The patent implements a self-service system where the machine learning recommender automatically performs the mapping task without requiring manual intervention. The system trains on ground truth datasets and then autonomously identifies relevant technical standard sections for patent claims, eliminating the need for manual review while maintaining high precision and significantly improving productivity.
Solution Approach 2:
The patent applies preliminary action by pre-training the machine learning model on ground truth datasets before actual patent processing. This preliminary training phase enables the system to learn optimal mapping patterns in advance, allowing it to quickly and accurately map new patent claims to relevant standard sections without manual intervention, thereby improving both precision and productivity.
3Measurement precision
If comprehensive ground truth datasets are used for training the machine learning algorithm, then recommendation accuracy improves, but data processing complexity and time increase
Solution Approach 1:
The patent segments the ground truth dataset into multiple components including claim text features, standard document features, technical domain classifications, and relevance labels. This segmentation allows the machine learning algorithm to process different aspects separately and combine them systematically, reducing overall processing complexity while maintaining high recommendation accuracy through structured feature engineering.
Data Source
AI summary
Machine learning based retrieval systems and methods are disclosed for mapping between patent or patent application claims and sections of technical standards Disclosed machine learning-based patent recommender systems may be trained and evaluated on example datasets obtained using the disclosed systems and methods. Systems and methods for generating ground truth datasets associating patent claims and sections of standards from information provided in intellectual property rights (IPR) disclosures or based on user feedback are also disclosed herein.