Research on Text Classification Methods Based on Decision Trees: A Case Study on the Recognition of the Entity Category 'Position'

Abstract

This paper investigates the application of a decision tree model for the binary classification task of the 'Position' category on the CLUENER2020 dataset, aiming to provide a lightweight and efficient method for named entity recognition. The CLUENER2020 dataset includes multiple label categories, among which the accurate identification of the 'Position' category is of significant importance for information extraction and text processing. Through data preprocessing, feature extraction, model training, and testing, this study evaluates the performance of the decision tree model on this task. The experimental results indicate that the model achieves an overall accuracy of 98%, with a precision of 98%, recall of 100%, and F1 score of 99% for the 'Non-Position' category, while the 'Position' category has a precision of 100%, recall of 85%, and F1 score of 92%. Although the model performs excellently on the 'Non-Position' category, the lower recall rate for the 'Position' category reveals a certain degree of missed detection, primarily attributed to the class imbalance in the dataset and the complexity of text features related to positions. The contribution of this paper lies in validating the applicability of traditional machine learning models for specific named entity recognition tasks. Particularly in resource-constrained scenarios, the decision tree model offers a feasible solution. Future research could further enhance model performance and improve the accuracy and robustness of named entity recognition tasks through data augmentation techniques, the integration of more complex model architectures, and in-depth feature engineering and hyperparameter optimization methods.