Enhanced Patent Classification using Natural Language Processing

The year 2023 saw an enormous rise in patent applications, with around 9700 patents filed daily, resulting in around 3.55 million annually. While this number saw a spike from last year’s statistics, the problem is the classification of patents. Currently, the process is not fully automated. There is some human intervention, which leads to delays in the process. With the help of this paper, we presented an approach using the techniques of Machine Learning and Natural Language Processing to classify these patents into their respective categories. In Natural Language Processing, we used methods like stopword removal, lowercasing, and vectorization. In contrast, Machine Learning is used to apply various models and test their accuracy and other performance metrics. The models used are Random Forest, Decision Tree, Support Vector, Logistic Regression, and Naïve Bayes, while the performance metrics are accuracy, precision, recall, F1-score, true positives, true negatives, false positives, and false negatives. These metrics are computed for each model, and the results are compared accordingly. Random Forest has outperformed all the models in all performance parameters. It has an accuracy of 0.94, precision of 0.9410, recall of 0.94, and F1-score of 0.9390. It also offers what further steps can be taken to improve the model’s performance. This research effort contributes significantly to classifying patents based on their aspects.