Research on patent classification based on hierarchical label semantics
Authors:
- 1 January 2022
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1025-1032
Abstract
Patent classification is an essential task in patent information management and knowledge mining. Most existing studies are based on the textual content of individual patent texts (e.g., titles and abstracts) for classification, but the patent also has label information with hierarchical structure and semantic description. However, the semantics correlation of patent texts and their labels have been largely ignored. In this paper, we propose a novel framework called HLSPC for patent classification by leveraging the hierarchical semantics correlation of patent texts and their labels. Specifically, we first apply a patent representation learning module for capturing the semantics representation of patent texts and hierarchical labels. Then, we design a label attention learning module to build the semantics correlation between patent texts and hierarchical labels, which enhances patent representation. Finally, we deploy a multi-level fusion module to get the refined category prediction for each patent which can preserve both local and global hierarchical prediction information. Extensive experimental results on two English patent datasets demonstrate the effective power of the HLSPC model.Keywords
Funding Information
- Research and Development
- National Natural Science Foundation of China
This publication has 23 references indexed in Scilit:
- Multilabel naïve Bayes classification considering label dependencePattern Recognition Letters, 2020
- Patent classification by fine-tuning BERT language modelWorld Patent Information, 2020
- Multi-label Patent Classification using Attention-Aware Deep Learning Model2020
- A deep learning analysis on question classification task using Word2vec representationsNeural Computing and Applications, 2020
- Hierarchical Multi-label Text Classification2019
- Hierarchical Transfer Learning for Multi-label Text Classification2019
- DeepPatent: patent classification with convolutional neural networks and word embeddingScientometrics, 2018
- Bag of Tricks for Efficient Text Classification2017
- Reduction strategies for hierarchical multi-label classification in protein function predictionBMC Bioinformatics, 2016
- Convolutional Neural Networks for Sentence Classification2014