Research on the dynamic application of trademark similarity determination based on dynamic time regularization and XGBoost

作者：Hong Ying Zhang, Xiaojuan Wang, Yaping Xu, Guijiang Hu · 年份：2023 · DOI：10.1145/3638584.3638612 · 被引用次数：1 · 研究领域：Medical Research and Treatments

In order to improve the possibility of successful trademark registration in the tobacco industry and monitor the problem of whether the 34 categories of trademarks are preempted, this paper conducts an experimental design based on dynamic time regularization and XGBoost algorithm. The main innovation points are firstly, the similarity is calculated by the Dynamic Time Wrangling (DTW) algorithm by combining the information of text structure and location, and also as a comparison with the accuracy of machine learning prediction results; secondly, both Chinese similarity and English similarity are incorporated into the training set as input features; thirdly, the weekly updated data training results are checked and used as the test set to enhance the accuracy and adaptability of the model. This experiment is applied to automatically output the determination of trademark similarity results, and then take measures to adjust the trademark design or raise objections to the infringing trademark after obtaining the similarity results, which greatly reduces the workload of manual comparison and improves the accuracy rate of comparison at the same time. The experimental design of this paper evaluates the accuracy rate, and the experimental results are in the acceptable range, which can be applied to the whole tobacco industry for screening similar brand names, and can be extended to other text similarity determinations, and this study can continue to optimize the extraction of more feat...