How Vital is the Jurisprudential Relevance: Law Article Intervened Legal Case Retrieval and Matching

TL;DR

This work addresses the challenge of jurisprudentially relevant legal case retrieval and matching (LCR/LCM) in civil-law contexts, where law articles encode essential reasoning beyond surface semantics. It proposes LCM-LAI, an end-to-end dependent multi-task framework that learns law article distributions through a law article prediction sub-task and couples semantic and legal interactions via an article-aware attention mechanism. Empirical results on four real-world datasets show state-of-the-art performance with strong efficiency and interpretability, demonstrating the practical viability of incorporating jurisprudential signals into case matching. The approach reduces reliance on expert annotations by leveraging law articles as intermediate variables, offering a scalable path to improved judicial support tools with broad applicability in civil-law systems.

Abstract

Legal case retrieval (LCR) aims to automatically scour for comparable legal cases based on a given query, which is crucial for offering relevant precedents to support the judgment in intelligent legal systems. Due to similar goals, it is often associated with a similar case matching (LCM) task. To address them, a daunting challenge is assessing the uniquely defined legal-rational similarity within the judicial domain, which distinctly deviates from the semantic similarities in general text retrieval. Past works either tagged domain-specific factors or incorporated reference laws to capture legal-rational information. However, their heavy reliance on expert or unrealistic assumptions restricts their practical applicability in real-world scenarios. In this paper, we propose an end-to-end model named LCM-LAI to solve the above challenges. Through meticulous theoretical analysis, LCM-LAI employs a dependent multi-task learning framework to capture legal-rational information within legal cases by a law article prediction (LAP) sub-task, without any additional assumptions in inference. Besides, LCM-LAI proposes an article-aware attention mechanism to evaluate the legal-rational similarity between across-case sentences based on law distribution, which is more effective than conventional semantic similarity. Weperform a series of exhaustive experiments including two different tasks involving four real-world datasets. Results demonstrate that LCM-LAI achieves state-of-the-art performance.

Figures (7)

• Figure 1: Example of the special text relevance in LCR. Left: the query and candidate legal cases in LCR. Right: two applicable law articles of the legal cases.
• Figure 2: The causal graph of two DMT frameworks. (a) The inconsistency between the training and reasoning stages leads to the train-test discrepancy. (b) The potential intermediate variable $R_L$ is introduced, making the training and reasoning process consistent.
• Figure 3: Overview of our framework LCM-LAI. This framework operates by taking textual descriptions of legal cases and text definitions of law articles as inputs. On the one hand, it leverages the Basic Interaction Module (BIM) to extract the semantic interaction information between cases based on the semantic similarity between across-case sentences. On the other hand, it uses the Legal Interaction Module (LIM) to capture the legal interaction information from the perspective of the similarity of legal distribution, in which the article prediction sub-task is introduced to capture the legal distribution of each sentence. Besides, LIM uses an article-intervened attention (AIA) mechanism to highlight the key jurisprudence-related parts of cases that rely on the predicted law articles. Finally, LCM-LAI combines the semantic and legal interaction representations to compute the matching score.
• Figure 4: Body Operations of the Legal Interaction Module. It cleverly exploits the intermediate representations of the attention mechanism of the law prediction sub-task. It not only takes the inner value vectors to compute the legal representations but also innovatively regards the attention score vectors of law prediction as the law article distribution for computing the legal-rational correlation between sentences across cases.
• Figure 5: The performance of LCM-LAI w.r.t. different mode of legal correlation matrix on LCM task.