基于多粒度语义交互的无监督法律裁判文书检索

Unsupervised Legal Case Retrieval Based on Multi-granularity Semantic-Aware Interaction

随着法律文书数据越来越多,信息过载问题日益严重,快速且准确地在海量法律文书中进行检索显得非常必要。法律文本作为一种特殊的文本形式,具有篇幅较长、结构复杂、专业性强等特点,传统基于关键字的文本检索方法不能满足用户查询法律信息的需求,容易出现答非所问、检索不全等问题。此外,基于语义的文本检索方法,大多依赖于对含有大量标注数据的法律文本进行有监督学习,而法律文本数据的人工标注则严重依赖专家知识,导致其需要高昂的人力成本。该文提出一种基于无监督学习的法律文书检索模型,分别从法律概念、词语和词组3个方面进行多粒度无监督文本匹配,避免了没有训练数据导致的冷启动问题。在法律裁判文书数据集上进行检索实验的结果表明,与基准模型相比,该模型在MAP、MRR和NDCG@10指标上均有显著提升,取得了优秀的检索效果,具有有效性和先进性。

With the ever-increasing size of legal cases in China,relevant legal case retrieval given a user query has attracted considerable attention.Conventional keyword-based retrieval systems look for matching cases that contain one or more words specified by the user.However,keyword searching is sharply focused on finding the exact terms specified in the query,making the retrieval systems miss many relevant documents.On the other hand,semantic-aware information retrieval methods usually rely heavily on labeled training data.Nevertheless,obtaining rich annotated data is a time-consuming and expensive process,creating a substantial barrier for applying supervised methods to legal case retrieval.This paper proposes an unsupervised,semantic-aware legal case retrieval method based on multi-granularity multigranularity semantic interaction.Specifically,leveraging legal concept-level features,phrase-level features,and word-level features to comprehensively explore the semantic-aware interaction between the query and each legal case.Experimental results on a real-life legal case retrieval corpus demonstrate that this method has a significant improvement in MAP,MRR and NDCG@10 indicators and substantially outperforms the baseline methods.