Mulan: A Multiple Residual Article-Wise Attention Network for Legal Judgment Prediction

Abstract

Legal judgment prediction (LJP) is used to predict judgment results based on the description of individual legal cases. In order to be more suitable for actual application scenarios in which the case has cited multiple articles and has multiple charges, we formulate legal judgment prediction as a multiple label learning problem and present a deep learning model that can effectively encode the content of each legal case via a multi-residual convolution neural network and the semantics of law articles via an article encoder. An article-wise attention mechanism is proposed to fuse the two types of encoded information. Experimental results derived on the CAIL2018 datasets show that our model provides a significant performance improvement over the existing neural models in predicting relevant law articles and charges.

KHAN: Knowledge-Aware Hierarchical Attention Networks for Political Stance Prediction

“…Political stance prediction. This work aims to solve the political stance prediction problem: given a news article 𝑎, predict its political stance (e.g., [1,5], where 1 indicates 'left' and 5 indicates 'right'). This problem, a typical supervised learning task, can be defined as follows from the optimization perspective:…”

“…Note that this 'sentence-wise' attention layer of HAN has advantages in terms of both efficiency and effectiveness, compared to the existing 'article-wise' attention methods [5,12] that take into account all words in the article together. HAN (1) requires the amount of computation much less than the existing methods (efficient) and ( 2) is able to effectively capture the local context without being interfered with the words that are located far away (effectiveness) because HAN only learns the relationships among closely located words.…”

Zero-shot Transfer of Article-aware Legal Outcome Classification for European Court of Human Rights Cases

“…Luo et al 2017 used an attention-based neural network which jointly models charge prediction and relevant article extraction in a unified framework whose input includes the text of legal articles. Sim-ilarly, Wang et al , 2019bChen et al 2022;Yue et al 2021 employ matching mechanism between case facts and article texts. To the best of our knowledge, ours is the first work to adapt articleaware prediction to the ECtHR corpus, which is situated in the in human rights litigation domain.…”

“…This work seeks to remedy this incomplete inference and enable the model to learn more authentic reasoning between rules and case facts by casting LJP into an article-aware classification setting and subjecting it to a zero-shot transfer challenge. Article-aware classification has been explored on Chinese criminal case corpora (Wang et al, , 2019bYue et al, 2021;Chen et al, 2022). Similarly, Holzenberger et al 2020 has modeled statutory reasoning by classifying US tax law provisions concatenated with textual case descriptions.…”

LPN: Label-Enhanced Prototypical Network for Legal Judgment Prediction

“…For example, Bao et al [12] proposed an attentional neural network, which predicted the legal article first and then employed the article to help improve charge prediction. Chen et al [13] proposed a multiple residual articlewise attention network, which adopted a multi-scale convolutional network to encode factual descriptions, and incorporated label information instead of just using labels as the index to achieve LJP. Considering the dependencies among multiple subtasks, Zhong et al [4] formalized the dependencies as a Directed Acyclic Graph(DAG) and proposed a topological multi-task learning framework to realize the three subtasks predictions, which is very representative and effective.…”