新住民相關親權酌定裁判書的文字探勘：對「平等」問題的法實證研究嘗試

Applying Texts Mining on Child Custody Cases Regarding New Immigrants in Taiwan: An Empirical Study on Equality

邵軒磊(Hsuan-Lei Shao) ； 黃詩淳(Sieh-Chuen Huang)

《臺大法學論叢》 49卷S期 (2020/11) Pp. 1267-1308

https://doi.org/10.6199/NTULJ.202011/SP_49.0001

摘要

本研究關心新住民與我國籍配偶離婚後，進入法院爭取未成年子女親權時，是否獲得平等對待之問題。本文使用資訊科學技術，驗證「法院是否會因父母的國籍而在裁判書中使用不同的詞彙」，分析116件親權酌定事件之裁判書。其中有31件之一方當事人為外籍、外裔人士（簡稱外籍裁判），85件為雙方當事人皆為我國籍（簡稱雙臺籍裁判）。首先，將裁判書中的詞彙在分詞後，轉化為TF-IDF的文本詞彙矩陣，再透過主成分分析投影在二維平面上，將每則裁判書的詞彙特徵（所在位置）可視化。雙臺籍裁判與外籍裁判並未分為不同的兩群，而是位置重合，意味著法院在兩種當事人的裁判並未使用不同的詞彙，因此無法直接主張司法體系對新住民不平等。其次，再使用機器學習中的梯度提升法，發現外籍裁判中，「是否出境」以及「居住狀況」相關詞彙特徵，較雙臺籍裁判為高，表示法官在外籍裁判中較關心程序參與以及居留問題。藉由文字探勘與機器學習來分析裁判書文本，除了檢驗平等問題外，還可將抽象的法律概念（子女最佳利益）具象化為裁判書中提及之事實（出境、租屋），供實務參考，結合了資訊與法律兩個領域，呼應本刊「跨域、科技與法律」的主題。

關鍵字

法律資料分析 ； 自然語言處理 ； 詞頻─倒數文件頻率法 ； 平等 ； 新住民 ； 法實證研究 ； 主成分分析

並列摘要

There are more and more transnational (international) marriage cases in Taiwan today. As an ethnic group, the New Immigrants are facing multiple issues regarding equality/inequality in daily life. This article attempts to find if a foreign spouse is treated equally in child custody proceedings in Taiwan. In detail, we studied 116 real legal texts (judgements) by Natural Language Processing (NLP) and Machine Learning (ML) in which 31 foreign cases are included. Firstly, after word segmentation, the document-term-matrix of TF-IDF is produced and then put into principle component analysis so that a two-dimensional plot is formed. It is thus found that 85 domestic marriage cases and 31 international marriage cases are almost located at the same place, meaning that the court does not use different vocabularies between these two types of cases. It indicates that the family judges did not treat or judge the New Immigrants differently from domestic couples. Secondly, using gradient boosting of machine learning, it is found that the vocabulary characteristics of cases involving New Immigrants are: ＂whether one party appears at court＂, ＂renting a house＂, and ＂living together＂, indicating that judges are more concerned about the participation of procedures and residence of New Immigrant parents. If a foreign spouse obtains legal aid and participates in the judicial process, since the standard adopted by the judge on the child custody is the same on both New Immigrant cases and domestic marriage cases, it is very possible that the foreign spouse will receive custody. However, the foreign spouse may need to acquire other social support on housing as it is a factor that family judges tend to investigate in New Immigrant cases. In this paper, the authors adopt disciplinarities of information science and law, showing one of the possibilities in which technology helps us to deepen the discussion of social responsibility.

並列關鍵字

Legal Analysis ； Natural Language Processing (NLP) ； TF-IDF ； Equality ； New Immigrants ； Empirical Legal Studies ； Principle Component Analysis

參考文獻

王章逸、闕河嘉（2020），〈大埔之歌：臺灣主流報紙中的「土地徵收」〉，《資訊社會研究》， 38 期 ， 頁 19-50 。http://doi.org/10.29843/JCCIS.202001_(38).0004

李震山（2004），〈憲法意義下之「家庭權」〉，《中正大學法學集刊》，16 期，頁 61-104。http://doi.org/10.30094/NCCULJ.200407.0002

林琬真、郭宗廷、張桐嘉、顏厥安、陳昭如、林守德（2012），〈利用機器學習於中文法律文件之標記、案件分類及量刑預測〉，《中文計算語言學期刊》，17 卷 4 期，頁 49-67。https://doi.org/10.30019/IJCLCLP.201212.0004

邵軒磊、吳國清（2019），〈法律資料分析與文字探勘：跨境毒品流動要素與結構研究〉，《問題與研究》 ， 58 卷 2 期，頁 91-114 。https://doi.org/10.30390/ISC.201906_58(2).0003

姜晴文（2019），《法律資料分析的優化與應用：以離婚後未成年子女親權酌定的裁判為素材》，國立臺灣大學法律學研究所碩士論文（未出版）。https://doi.org/10.6342/NTU201904293