基于结构内容特征的裁判文书自动推荐研究

Automatic Recommendation of Judgment Documents Based on Structural Content Features

在现有的裁判文书检索系统上,非专业领域的用户检索具有局限性。目前,法律领域的智能检索仅在基于裁判文书的法律条文的推荐和分类上开展了研究,缺乏对裁判文书自动推荐的相关研究,因此,本文提出了一种利用类新闻的事实性文本智能推荐裁判文书的方法,结合目前的研究工作,总结裁判文书的结构和内容特征,利用类新闻的事实性文本模拟非法律专业用户的检索查询式,构建含有结构内容特征的裁判文书语料库,并自动推荐相关裁判文书文档。结果显示,利用裁判文书的法院意见结构内容特征,对新闻语料进行特征词表示之后,LambdaMART模型在文本匹配结果上表现良好,优于传统的全文检索技术。

In the existing judgment document retrieval system,non-legal professionals encounter difficulties while trying to retrieve documents.At present,intelligent searches in the legal field can rely on the recommendations and classifications of legal provisions based on judgment documents without researching the automatic recommendations of judgment docu-ments.Therefore,this study proposes a method to intelligently recommend judgment documents using news texts based on similar news methods.This article combines current research work to summarize the structure-specific and content charac-teristics of judgment documents,use the text of news articles,simulate the search query of non-legal professional users,and construct a corpus of judgment documents with structural content features to recommend automatically related judg-ment documents.The results show that after extracting the corresponding feature words using the content structure of the court opinion of judgment documents,the retrieval model performs well when the LambdaMART model conducts text matching,which is superior to the traditional full-text retrieval technology.