Few-Shot Charge Prediction with Discriminative Legal Attributes

Proceedings of the 27th International Conference on Computational Linguistics, pages 487–498 Santa Fe, New Mexico, USA, August 20-26, 2018. 487 Few-Shot Charge Prediction with Discriminative Legal Attributes Zikun Hu∗ Xiang Li∗ Cunchao Tu Zhiyuan Liu† Maosong Sun Department of Computer Science and Technology, Tsinghua University State Key Lab on Intelligent Technology and Systems, Tsinghua University Beijing National Research Center for Information Science and Technology {hzk14, x-l15}@mails.tsinghua.edu.cn tucunchao@gmail.com, {liuzy, sms}@tsinghua.edu.cn Abstract Automatic charge prediction aims to predict the final charges according to the fact descriptions in criminal cases and plays a crucial role in legal assistant systems. Existing works on charge prediction perform adequately on those high-frequency charges but are not yet capable of pre- dicting few-shot charges with limited cases. Moreover, these exist many confusing charge pairs, whose fact descriptions are fairly similar to each other. To address these issues, we introduce several discriminative attributes of charges as the internal mapping between fact descriptions and charges. These attributes provide additional information for few-shot charges, as well as effective signals for distinguishing confusing charges. More specifically, we propose an attribute-attentive charge prediction model to infer the attributes and charges simultaneously. Experimental results on real-work datasets demonstrate that our proposed model achieves significant and consistent improvements than other state-of-the-art baselines. Specifically, our model outperforms other baselines by more than 50% in the few-shot scenario. Our codes and datasets can be obtained from https://github.com/thunlp/attribute_charge. 1 Introduction The task of automatic charge prediction aims to train a machine judge to determine the final charges (e.g., theft, robbery or traffic offence.) of the defendants in criminal cases. As a representative subtask of legal judgment prediction, charge prediction plays an important role in legal assistant systems and can benefit many real-world applications. For example, it can provide a handy reference for legal experts (e.g., lawyers and judges) and improve their working efficiency. Meanwhile, it can supply ordinary people who are unfamiliar with legal terminology and complex procedures with legal consulting. As a typical task in legal intelligence, automatic charge prediction has been studied for decades and most existing works formalize this task under the text classification framework. At the early stage, researchers pay great efforts to extract efficient features from text or case profiles. For example, some works (Liu et al., 2004; Liu and Hsieh, 2006) utilize shallow textual features, including characters, words, and phrases, to predict charges. Katz et al. (2017) predict the US Supreme Court’s decisions with efficient features extracted from case profiles (e.g., dates, locations, terms, and types). All these approaches require numerous human effort to design features and annotate training instances. Besides, these methods are hard to scale to other scenarios. Inspired by the successful usage of deep neural networks on natural language processing tasks (Kim, 2014; Baharudin et al., 2010; Tang et al., 2015), researchers propose to employ deep neural networks to model legal documents. For example, Luo et al. (2017) propose an attention-based neural network for charge prediction by incorporating the relevant law articles. However, charge prediction is still confronted …

Content truncated to 500 words