請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/20199
標題: | 基於法律判決書之公司倒閉風險評估 Company Failure Risk Assessment Based on the Court’s Judgments |
作者: | CHIH-YANG HUANG 黃志揚 |
指導教授: | 項潔(Jieh Hsiang) |
關鍵字: | 自然語言處理,法律文件,風險分析,線性分類器,長短期記憶, natural language processing,judgment document,risk analysis,linear classification,long short-term memory, |
出版年 : | 2021 |
學位: | 碩士 |
摘要: | 法律是一整套形式化、意義明確的法規條文,當今社會皆依其規範運作。人們的社會活動中,若有糾紛或利益衝突時,便會尋求法官根據法律進行判決,最終產出法律判決書,法律可謂反映著當今社會的人類活動與社會狀態。法院判決行之有年,累積的判例卷帙浩繁,龐大的資料量迄今已超出人力處理得以企及的程度;因此,以電腦輔助處理,在現代已是大勢所趨,本研究將聚焦討論「公司經歷的法律案件」對於「公司是否續存」的影響。公司行號經營事業,時常面臨各式訴訟,尤其在公司經營狀況不佳時,隨之而來的法律訴訟,例如:債務違約、侵權、掏空資產等,對於公司行號的存續與否將是相當大的考驗,而法律判決書,往往可以一窺公司的經營問題。在判決過後,法院會出具法律判決書,以做為訴訟的紀錄。判決書中包含了兩造雙方、案由、判決理由、最終判決等內容,為資訊含量極高的文件。綜觀過去基於人工智慧的法律研究,多半聚焦於判決類型的分類和刑期的預測,對於公司行號的經營風險則甚少著墨。有鑑於此,本論文將自然語言的處理技術應用於法律領域,剖析法律判決書,透過歷年的判決資料,結合公司營運狀況,試圖建立模型,找出公司倒閉風險與法律判決書之間的關聯,使研究者得以迅速掌握公司風險概況。其次,我們也嘗試在判決書全文中找出一些有意義的文字,並 分析公司行號的營運風險與這些詞彙之間的關係。本研究將自然語言處理的常用技術應用於法律文本上,並比較了幾個不同的機器學習模型,以期許能為公司經營與法律的分析提供另一條不同的研究取徑。 Law, according to which people are living, is a system regulating the present society with specific, formal, and clear codes. If a conflict or dispute occurs in the activity of society, people often resort to the court to have their problem settled by the law, and the result of the judgment is named verdict. Through verdicts, we are able to get a quick grasp of the status and activity of the society.So far, there are plentiful verdicts, and such abundance makes manual search and analysis almost impractical. However, with the assistance of computers, people could deal with that easily, which is also the trending procedure in the world. In this research, we will focus on detection of company failure risks from historical verdicts of a company.A company could be accused of various things, especially when the management is poor. For example, a company would face debt default, infringement or hollowed out assets, to name a few, and such issues are critical for the operation of a company. However, based on past verdicts of a company, we could peek the operation status of a company. A verdict is a record of judgment issued by the court after the trial. A verdict, as an informative document, is composed of litigants, cause of the case, reason and the final decision, etc. So far, most previous researches on verdict analysis using artificial intelligence focus on judgment classification and sentence prediction, but there is rather few focusing on company failure risks. In the light of this, in this thesis, we will apply natural language processing skills on the field of law to handle the verdicts, and integrate past verdicts and the business status of a company to assess the company failure risk and to accelerate the process of understanding the risk of a company. Next, we recognize some interesting and meaningful terms in the verdicts, discovering the relation between a company and these terms.We use some common natural language processing skills on the judgment documents, comparing the different machine learning models as an expectation of providing another novel research approach. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/20199 |
DOI: | 10.6342/NTU202004222 |
全文授權: | 未授權 |
顯示於系所單位: | 資訊工程學系 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
U0001-2009202020205100.pdf 目前未授權公開取用 | 7.61 MB | Adobe PDF |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。