Categorization of Hand Gesture Recognition Patents with PatSeer’s AI Classifier: A Case Study
Quick and accurate patent categorization is important because it enables you to comprehend the IP landscape of an industry. Manual categorization is tedious and takes hours. Occasionally it’s not possible to do because of insufficient subject matter experts. IPC and CPC classes are wide-ranging and might not match your categories. With these considerations in mind, we created PatSeer’s AI Classifier to reduce the amount of classification that has to be done constantly for technologies that you’re keeping tabs on.
Built over the latest AI & NLP stack, the underlying Classifier model has been customized with semantic rules in ReleSense™ to achieve greater precision over patent and scientific literature. The AI Classifier can automatically learn from your existing taxonomy and the records you have categorized against it to make predictions for newer, uncategorized records. It supports hierarchical taxonomies and saves you the time and effort of manually classifying new records. The Classifier once enabled on your projects continues to learn in the background based on your feedback. Multiple modes of operation allow you to choose the level of intervention you want in the process.
This case study evaluates PatSeer’s AI Classifier using a manually classified dataset on hand gesture recognition patents.
Dividing the dataset for training and testing purposes
Since hand gesture recognition has a large number of applications, to evaluate the Classifier a subset of the matching patents viz. a total of 200 records were taken into consideration. For the training dataset, we manually categorized 60 records (15 for each category) into Gaming, TV, Sign Language, and Medical Applications. The AI Classifier built a prediction model over the training dataset and categorized the remaining 140 records. The prediction threshold (the value at which the decision is made whether to classify a patent under the selected categories or not) was set to 60 for our test purposes.
Read more: The assessment of the AI Classifier’s performance based on 4 machine learning (ML) performance metrics.
