효과적인 전자증거개시 처리를 위한 질의 추천 시스템

Abstract

In order to effectively respond to the future litigation and to win a case, the most important task is to secure the highly relevant evidence produced by the proper search and review in respect of litigation issues during a whole e-Discovery procedure. At this point, the role of search is to reduce the amounts of document which should be additionally reviewed as potential evidence, so the poor search result caused by the use of wrong keywords can bring unexpected time and cost problems. These keywords, in general, are selected by analysis about the content of complaint or related documents at the beginning stage of e-Discovery and this stage is called ECA(Early Case Assessment) in EDRM(Electronic Discovery Reference Model). Ultimately, the success of e-Discovery depends very much on how well the participants performed essential tasks of ECA, but it has mostly depended on the ability of specific human like lawyer because existing e-Discovery solutions did not support this kind of function in priority. This thesis, therefore, suggests the machine learning based litigation preparing method for effective early case assessment on e-Discovery procedure. The suggested method extracts and collects the meaningful information from the complaints and related documents which were retained by the litigant. This information can be used for identifying the main issues of litigation, discussion in meet-and-confer session, creating a request for production, or writing another related complaint. Also, special experiment and evaluation result using the real complaint introduced by TREC Legal Track will be described for analyzing the availability of this method.