基于Text-CNN联合分类与匹配的合同法律智能问答系统研究
2020-06-21刘葛泓李金泽李卞婷邵南青窦万峰
刘葛泓 李金泽 李卞婷 邵南青 窦万峰



摘 要:面向法律领域的相关问题,需要借助专业的法律文本。利用司法领域的文本资源解决用户提出的合同法律问题,能在很大程度上降低人工成本,节约社会资源。为了更加智能、高效地响应用户在合同法方面的法律诉求问题,本文设计与实现了一个合同法律智能问答系统,并给出了一种基于文本卷积神经网络(Text-CNN)的联合分类与匹配的合同法律智能问答深度学习模式,针对合同法领域的文本特征,对其进行了分类。实验表明,该模式适合于合同法领域的智能问答。
关键词:合同法律智能问答系统;文本卷积神经网络;自然语言处理;词向量
Abstract: Relevant issues in the legal field require professional legal texts. The text resources in the judicial field can be used to deal with the legal questions concerning contracts raised by users, which can greatly reduce labor cost and save social resources. In order to more intelligently and efficiently respond to users' legal claims in contract law, this paper designs and implements an intelligent contract law Question and Answer System (QAS), and proposes a deep learning model of intelligent QAS focusing on contract law based on the technology of joint classification and matching of Text Convolutional Neural Network (Text-CNN), which can classify different texts of contract law according to their characteristics. Experiments show that this model is suitable for intelligent question answering in the field of contract law.
Keywords: intelligent contract law question-and-answer system; Text-CNN; natural language processing; word vector convolutional neural network;natural language processing;the word vector
1 引言(Introduction)
隨着我国市场经济的发展,合同纠纷成为人们生活中最常出现的法律问题之一。因此,精准高效地获取法律援助成为每个合同主体的诉求。随着互联网的发展,传统的搜索引擎已不能满足用户的需求,智能问答系统应运而生。问答系统的核心是文本匹配[1]。虽然Prolo[2]、Monroy等人[3]在该领域有一定贡献,但一定程度上依赖于人工标注、答案范围有限。同时,针对合同领域问答系统的相关研究较少,未有较好的解决方案。
针对上述问题,本文通过词向量技术构建合同法语料库,采用联合加分类的方式,利用文本卷积神经网络(Text-CNN[4])模型对语料库进行特征分类,划分问题类别,缩小答案映射范围。经实验,Text-CNN文本分类模型适合于问题特征复杂且多样的合同法律问答系统,相较于LSTM[5]模型在文本分类上拥有更高的效率。……
