SOTAVerified

Patent classification

Patent reviewers usually are responsible to classify the patent applications, i.e., they assign the design codes. This is time-consuming due to the numerous classification codes. For instance, the U.S. design patent system has 33 classes which are further divided into subclasses. Given that design patents include both titles and visual content, the goal of the experiment is patent classification by integrating titles, captions, and images. Although a single design patent can be associated with multiple design codes, we focus on the primary classification—solving the task as a multi-class classification problem.

Papers

Showing 26–36 of 36 papers

TitleStatusHype
ClusterDataSplit: Exploring Challenging Clustering-Based Data Splits for Model Performance EvaluationCode0
Deep learning-based citation recommendation system for patents—0
A Convolutional Neural Network-based Patent Image Retrieval Method for Design Ideation—0
BERT-CNN: a Hierarchical Patent Classifier Based on a Pre-Trained Language Model—0
A Label Informative Wide \& Deep Classifier for Patents and Papers—0
PatentBERT: Patent Classification with Fine-Tuning a pre-trained BERT ModelCode0
Universal Language Model Fine-tuning for Patent Classification—0
Classifying Patent Applications with Ensemble Methods—0
Long-run dynamics of the U.S. patent classification system—0
Filtering Patent Maps for Visualization of Diversification Paths of Inventors and Organizations—0
Text Representations for Patent Classification—0
Show:102550
← PrevPage 2 of 2Next →

No leaderboard results yet.