Few-Shot Charge Prediction with Data Augmentation and Feature Augmentation

Abstract

The task of charge prediction is to predict the charge based on the fact description. Existing methods have a good effect on the prediction of high-frequency charges, but the prediction of low-frequency charges is still a challenge. Moreover, there exist some confusing charges that have relatively similar fact descriptions, which can be easily misjudged. Therefore, we propose a model with data augmentation and feature augmentation for few-shot charge prediction. Specifically, the model takes the text description as the input and uses the Mixup method to generate virtual samples for data augmentation. Then, the charge information heterogeneous graph is introduced, and a novel graph convolutional network is designed to extract distinguishability features for feature augmentation. A feature fusion network is used to effectively integrate the charge graph knowledge into the fact to learn semantic-enhanced fact representation. Finally, the semantic-enhanced fact representation is used to predict the charge. In addition, based on the distribution of each charge, a category prior loss function is designed to increase the contribution of low-frequency charges to the model optimization. The experimental results on real-work datasets prove the effectiveness and robustness of the proposed model.

Methods and Applications of Data Mining in Business Domains

“…They offer an advanced model for detecting financial statement fraud using XGBoost [12], introduce innovative neural network models for stock price prediction [13], analyze factors influencing tourist offer prices [14], develop predictive models for healthcare patient influx [15], and propose intelligent decision forest models for customer churn prediction in the telecom industry [16]. They also address customer churn prediction in noncontractual B2B settings [17], improve legal judgment prediction through graph neural networks [18], enhance car sales forecasts using online sentiment data and deep learning [19], introduce a reinforcement learning framework for options trading [20], and predict the charge of a legal case using a novel graph convolutional network [21]. These studies showcase the versatility and practical applications of data-driven techniques in diverse fields, underscoring their importance for informed decision making and predictive accuracy.…”

LPN: Label-Enhanced Prototypical Network for Legal Judgment Prediction

“…Hu et al [5] achieved few-shot charge prediction by introducing some attributes constructed manually, but it has poor transferability. And Han et al [16] proposed BERT-Attention based on easy data augmentation techniques to achieve fewshot charge prediction, which added a large amount of data to alleviate the imbalanced distribution of the original data, but it is easy to cause overfitting. Recently, Liu et al [6] constructed case triples as input which contain two similar cases and one dissimilar case, using the relationship and frequency information of cases to optimize model learning, and achieved good results.…”