摘要
预测编码是一种利用计算机技术审查、分析和处理大量文档的技术。预测编码在特定算法支持下进行"文本分类",并基于律师及案情专家对测试文件审阅,通过软件程序对与特定案件或争论点相关或者可响应的文件进行识别。由于预测编码可以通过协助律师进行文件的搜索、挑选和分类,减少需要手动审查的不相关和无响应文档的数量从而大幅降低电子披露流程的成本,被认为是电子证据开示领域最重要的发展之一。自2012年以来,在英美法系中已有若干在司法审判中使用预测编码的案例。但由于预测编码技术仍处于发展之中,其仍然具有局限性。预测编码技术尚难以准确无误地捕捉到案件文本的关键点,在算法决策透明度、训练模型的设计上也仍不成熟。由于预测编码技术自身的专业性,使得法律文件审查工作将日益专业化和精英化。人类语言的复杂性使得预测编码尚难以准确理解不同场景中不同词语的内涵,因而在审查的准确性上仍存在风险。
Predictive coding is the use of keyword search,filtering and sampling to automate portions of an e-discovery document review.It is enable to realize'text categorization'under the application of specific algorithms and identify critical issues by softwares.It has been regarded as one of the most important developments in electronic evidence discovery area.Since 2012,there exist several judicial cases that applying predictive coding in common law countries.Nevertheless,there are several shortcomings in predictive coding.It is still hard to accurately predict key points of case documents.There exist weaknesses in the transparency of algorithms decision-making process and its model design.It is also difficult to accurately understand complicated human language under different scenarios.
出处
《经贸法律评论》
2018年第1期97-103,共7页
Business and Economic Law Review
关键词
预测编码
电子证据开示
人工智能
Predictive Coding
E-Evidence Discovery
Artificial Intelligence
