臺灣博碩士論文加值系統

臺灣的法律雖日益成熟，但對非專業人士來說，比起理解法律本身，尋找與自身情況相似的裁判書並從中得知可能相關的法律和處理程序還更加有用。然而現有的裁判書查詢系統僅支援關鍵詞搜尋，無法與使用者的案件情境 ( Case Scenario ) 有所連結。因此，本論文旨在改善現有系統，使社會大眾可以透過案件情境搜尋相關法條和裁判書，獲得初步法律專業知識，以便進行後續相關流程。本論文使用司法院法學資料檢索系統內的裁判書資料，以基於變換器的雙向編碼器表示技術 ( Bidirectional Encoder Representations from Transformers , 以下簡稱 BERT ) 作為主要研究方法。首先以資料收集及前處理模組 ( Data Collection and Pre-processing Module, DCPM ) 進行資料的收集和處理。本論文先著重於裁判書案件情境的抓取，於裁判書內案件情境偵測模型建置模組 ( Construction of Judgement-based Case Scenarios Detection Model Module, CJCSDMM ) 內建立兩階段式案件情境擷取模型，幫助完成整體案件情境資料集的建置。接著本論文於刑事相關法規及內部法條預測模型建置模組 ( Construction of Criminal Law and Internal Article Prediction Model Module, CCLIAPMM ) 建立能夠預測法規和法條的模型。最後以聊天機器人應用模組 ( Chatbot Application Module, CAM ) 開發聊天機器人，透過使用者的案件情境以及預測的法規和法條結果，尋找與使用者的案件情境狀況大致相符的裁判書。本系統透過簡單直覺操作，希望能貼近使用者的生活，減少人們與法律之間的隔閡。

Although the legal system in Taiwan is becoming more sophisticated, for non-professionals, it is more helpful to search for judgments that are similar to their situation and learn about the possible relevant laws and procedures rather than simply understanding the law itself. However, existing judgment retrieval systems only support keyword searches and cannot be linked to the user's case scenarios. Therefore, this paper aims to improve the existing system so that the general public can search for relevant laws and judgments through case scenarios, obtain prior legal knowledge, and proceed with subsequent processing. This paper uses the judgments data in the Judicial Yuan Legal Information Retrieval System and the Bidirectional Encoder Representations from Transformers (BERT) as the primary research method. First, the Data Collection and Pre-processing Module (DCPM) is used for data collection and processing. This paper focuses on extracting case scenarios from judgments. It establishes a two-stage case scenario extraction model in the Construction of Judgement-based Case Scenarios Detection Model Module (CJCSDMM) to help complete the overall case scenario dataset. Then, the Construction of Criminal Law and Internal Article Prediction Model Module (CCLIAPMM) is used to establish models that can predict laws and articles. Finally, a Chatbot Spplication Module (CAM) is developed to find judgments roughly consistent with the user's case scenario based on the user's case scenario and the predicted laws and articles. This system hopes to be closer to the user's life through simple and intuitive operations and reduce the gap between people and the law.

摘要...........................................................................i
Abstract......................................................................ii
誌謝.........................................................................iii
目錄..........................................................................iv
表目錄......................................................................viii
圖目錄.........................................................................x
第一章 緒論...................................................................1
1.1 研究背景...............................................................1
1.2 研究動機...............................................................2
1.3 研究目的...............................................................3
第二章 研究技術探討............................................................4
2.1 裁判書.................................................................4
2.2 聊天機器人.............................................................5
2.3 自然語言處理 ( Natural Language Processing , NLP )......................6
2.3.1 斷詞...................................................................6
2.3.2 詞向量.................................................................7
2.4 Gensim.................................................................8
2.5 詞頻-逆向文本頻率指數 ( Term Frequency-Inverse Document Frequency , TF-IDF ) .......................................................................8
2.6 TextRank...............................................................9
2.7 稀疏矩陣 ( Sparse Matrix ).............................................9
2.8 餘弦相似度............................................................10
2.9 Python Flask..........................................................10
2.10 Jieba.................................................................11
2.11 中文斷詞暨實體辨識系統 ( Chinese Knowledge And Information Processing Tagger, CKIP Tagger ).........................................................11
2.12 基於變換器的雙向編碼器表示技術 ( Bidirectional Encoder Representations from Transformers , BERT )....................................................11
2.12.1 Transformer...........................................................12
2.12.2 Encoder...............................................................12
2.12.3 自注意力機制 ( Self-Attention )........................................13
2.13 Longformer BERT.......................................................14
2.14 Gradient Checkpointing................................................16
2.15 分類器 ( Classifier ).................................................17
2.15.1 監督式學習............................................................17
2.15.2 非監督式學習...........................................................18
2.15.3 半監督式學習...........................................................18
2.15.4 二元 / 多元分類 ( Binary / Multi-Class Classification )................19
2.15.5 多標籤分類 ( Multi-Label Classification ).............................20
2.16 混淆矩陣 ( Confusion Matrix ).........................................23
2.16.1 Accuracy..............................................................25
2.16.2 Precision.............................................................25
2.16.3 Recall................................................................26
2.16.4 F1-Score..............................................................26
2.17 接收者操作特徵曲線 ( Receiver Operating Characteristic Curve , ROC Curve ).............................................................................27
2.18 潛在語意索引 ( Latent Semantic Indexing , LSI )........................28
2.19 相關研究..............................................................29
第三章 研究方法與步驟.........................................................30
3.1 資料收集及前處理模組...................................................31
3.1.1 資料擷取..............................................................32
3.1.2 段落分割..............................................................33
3.1.3 資料前處理............................................................34
3.2 裁判書內案件情境偵測模型建置模組........................................35
3.2.1 資料標註..............................................................36
3.2.2 兩階段式預測模型訓練...................................................39
3.3 刑事相關法規及內部法條預測模型建置模組...................................41
3.4 聊天機器人應用模組.....................................................44
3.4.1 案件情境相關法規條文查詢................................................45
3.4.2 刑事相關法規解釋查詢...................................................50
3.4.3 其他功能..............................................................50
第四章 研究結果..............................................................51
4.1 裁判書案件情境段落偵測模型..............................................51
4.2 裁判書案件情境偵測模型..................................................53
4.3 案件情境萃取系統流程...................................................54
4.4 案件情境之刑事相關法規預測模型結果.......................................55
4.5 案件情境之刑事相關法規內部法條預測模型結果...............................56
4.6 聊天機器人應用介面.....................................................58
4.6.1 案件情境相關法規條文查詢................................................59
4.6.2 刑事相關法規解釋查詢...................................................66
4.6.3 其他功能..............................................................67
第五章 結論..................................................................69
第六章 未來展望..............................................................70
參考文獻......................................................................71
附錄..........................................................................74
本論文收錄三十種刑事相關法規內部法條預測模型......................................74
1. 刑事訴訟法............................................................74
2. 中華民國刑法...........................................................75
3. 中華民國刑法施行法.....................................................76
4. 毒品危害防制條例.......................................................77
5. 洗錢防制法............................................................78
6. 藥事法................................................................79
7. 槍砲彈藥刀械管制條例...................................................80
8. 組織犯罪防制條例.......................................................81
9. 道路交通管理處罰條例...................................................82
10. 家庭暴力防治法.........................................................83
11. 陸海空軍刑法...........................................................84
12. 貪污治罪條例...........................................................85
13. 刑事妥速審判法.........................................................86
14. 通訊保障及監察法.......................................................87
15. 性侵害犯罪防治法.......................................................88
16. 廢棄物清理法...........................................................89
17. 懲治走私條例...........................................................90
18. 管制藥品管理條例.......................................................91
19. 森林法................................................................92
20. 醫師法................................................................93
21. 刑事訴訟法施行法.......................................................94
22. 警察職權行使法.........................................................95
23. 軍事審判法............................................................96
24. 水土保持法............................................................97
25. 證人保護法............................................................98
26. 兒童及少年性剝削防制條例................................................99
27. 社會秩序維護法........................................................100
28. 保安處分執行法........................................................101
29. 少年事件處理法........................................................102
30. 性騷擾防治法..........................................................103
Extended Abstract............................................................104
I. INTRODUCTION.........................................................105
II. BACKGROUND...........................................................105
A. Bidirectional Encoder Representations from Transformers..............105
B. Term Frequency-Inverse Document Frequency............................105
C. Latent Semantic Indexing.............................................105
III. METHOD...............................................................106
A. Data Collection and Pre-processing Module, DCPM......................107
B. Construction of Judgement-based Case Scenarios Detection Model Module, CJCSDMM......................................................................108
C. Construction of Criminal Law and Internal Article Prediction Model Module, CCLIAPMM.............................................................110
D. Chatbot Application Module, CAM......................................111
IV. Result...............................................................112
A. Judgment Case Scenarios Detection Model..............................112
B. Judgment Case Scenarios Paragraph Detection Model....................112
C. Criminal-Related Law Prediction Model................................113
D. Each Law Internal Article Prediction Model...........................113
E. ChatBot Demo.........................................................115
V. CONCLUSION...........................................................116
VI. REFERENCE............................................................117

[1]Wikipedia. 中華民國法律. from https://zh.wikipedia.org/zh-tw/%E4%B8%AD%E8%8F%AF%E6%B0%91%E5%9C%8B%E6%B3%95%E5%BE%8B
[2]中央研究院-詞庫小組. (1986). CHINESE KNOWLEDGE AND INFORMATION PROCESSING. from https://ckip.iis.sinica.edu.tw/
[3] 王安定（2016）。判決書之探勘分析與量刑迴歸模型之建立~以法院毒品判決書為例。未出版之碩士論文，臺北市立大學資訊科學系碩士在職專班，臺北市。
[4]司法院. 民事訴訟. from https://www.judicial.gov.tw/tw/cp-1447-58182-7cb42-1.html
[5]司法院法學資料檢索系統（）。裁判書查詢系統。 取自 https://law.judicial.gov.tw/FJUD/Default_AD.aspx
[6]全國法規資料庫（）。刑事訴訟法 第 308 條。 取自 https://law.moj.gov.tw/LawClass/LawSingle.aspx?pcode=C0010001&flno=308
[7]全國法規資料庫（）。法院組織法 第 83 條。 取自 https://law.moj.gov.tw/LawClass/LawSingle.aspx?pcode=A0010053&flno=83
[8]朱璟軒（2021）。一個基於深度學習之刑度預測模型—以臺灣地區竊盜案件為例。未出版之碩士論文，國立臺灣科技大學資訊工程系，台北市。
[9]吳晨皓（2020）。BERT與GPT-2分別應用於刑事案件之罪名分類及判決書生成。未出版之碩士論文，國立高雄科技大學資訊管理系，高雄市。
[10]林婷嫻. (2018). 斷開中文的鎖鍊！自然語言處理 (NLP)是什麼？. Retrieved from https://research.sinica.edu.tw/nlp-natural-language-processing-chinese-knowledge-information/
[11]紀幸辰（2018）。一個基於機器學習的醫療法判決書預測系統使用具文本相似性的法條分類決策樹。未出版之碩士論文，國立臺北大學資訊工程學系，新北市。
[12]高點法商編委會（2022）。高點體系式分類六法：刑事法規(含法律倫理)）：高點文化事業有限公司。
[13]曹錫璋（2021）。基於深度學習模型之判決書情境相似檢索技術之研究。未出版之碩士論文，國立中興大學資訊科學與工程學系所，台中市。
[14]陳冠瑜（2015）。利用語意分析模型分析谷歌部落格搜尋引擎效能。未出版之碩士論文，國立東華大學資訊管理碩士學位學程，花蓮縣。
[15]曾紀秀（2021）。酒駕犯罪判決書之大數據採礦分析。未出版之碩士論文，國立中正大學雲端計算與物聯網數位學習碩士在職專班，嘉義縣。
[16]匯澤法律事務所. 刑事案件. from https://huitselaw.com/service/list/Criminal-case/
[17]楊舒婷（2020）。怎麼看懂判決書？判決書上都記載了什麼內容？（二）——刑事判決書。 取自 https://www.legis-pedia.com/article/lawABC/739
[18] 謝閎宇（2021）。基於深度學習模型之刑事判決書情境萃取研究。未出版之碩士論文，國立中興大學資訊科學與工程學系所，台中市。
[19]謝德成（2021）。以機器學習為基礎之車禍致死案件精神損害賠償判決預測。未出版之碩士論文，國立暨南國際大學資訊工程學系，南投縣。
[20]Bafna, P., Pramod, D., & Vaidya, A. (2016). Document clustering: TF-IDF approach. Paper presented at the 2016 International Conference on Electrical, Electronics, and Optimization Techniques (ICEEOT).
[21]Beltagy, I., Peters, M. E., & Cohan, A. J. a. p. a. (2020). Longformer: The long-document transformer.
[22]Bradley, A. P. J. P. r. (1997). The use of the area under the ROC curve in the evaluation of machine learning algorithms. 30(7), 1145-1159.
[23]Brand, M. J. L. a., & applications, i. (2006). Fast low-rank modifications of the thin singular value decomposition. 415(1), 20-30.
[24]Chen, T., Xu, B., Zhang, C., & Guestrin, C. J. a. p. a. (2016). Training deep nets with sublinear memory cost.
[25]Cui, Y., Che, W., Liu, T., Qin, B., Yang, Z. J. I. A. T. o. A., Speech,, & Processing, L. (2021). Pre-training with whole word masking for chinese bert. 29, 3504-3514.
[26]Cui, Y., Che, W., Liu, T., Qin, B., Wang, S., & Hu, G. J. a. p. a. (2020). Revisiting pre-trained models for Chinese natural language processing.
[27]Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. J. a. p. a. (2018). Bert: Pre-training of deep bidirectional transformers for language understanding.
[28]Giri, R. K., Gupta, S. C., & Gupta, U. K. (2021). An approach to detect offence in memes using natural language processing (NLP) and deep learning. Paper presented at the 2021 International Conference on Computer Communication and Informatics (ICCCI).
[29]Heydarian, M., Doyle, T. E., & Samavi, R. J. I. A. (2022). MLCM: Multi-label confusion matrix. 10, 19083-19095.
[30]Hofmann, T. (1999). Probabilistic latent semantic indexing. Paper presented at the Proceedings of the 22nd annual international ACM SIGIR conference on Research and development in information retrieval.
[31]Is there a method for calculating the best accuracy threshold for multi-label classification? (Lesson 3). (2019). from https://forums.fast.ai/t/is-there-a-method-for-calculating-the-best-accuracy-threshold-for-multi-label-classification-lesson-3/75330
[32]Lan, Z., Chen, M., Goodman, S., Gimpel, K., Sharma, P., & Soricut, R. J. a. p. a. (2019). Albert: A lite bert for self-supervised learning of language representations.
[33]Li, P.-H., Fu, T.-J., & Ma, W.-Y. (2020). Why attention? Analyze BiLSTM deficiency and its remedies in the case of NER. Paper presented at the Proceedings of the AAAI Conference on Artificial Intelligence.
[34]Liddy, E. D. (2001). Natural language processing.
[35]Liu, Y., Ott, M., Goyal, N., Du, J., Joshi, M., Chen, D., . . . Stoyanov, V. J. a. p. a. (2019). Roberta: A robustly optimized bert pretraining approach.
[36]Mihalcea, R., & Tarau, P. (2004). Textrank: Bringing order into text. Paper presented at the Proceedings of the 2004 conference on empirical methods in natural language processing.
[37]Mikolov, T., Chen, K., Corrado, G., & Dean, J. J. a. p. a. (2013). Efficient estimation of word representations in vector space.
[38]O'Shea, K., & Nash, R. J. a. p. a. (2015). An introduction to convolutional neural networks.
[39]OpenAI. (2022). Chat Generative Pre-trained Transformer,ChatGPT. from https://chat.openai.com/chat
[40]Seliya, N., Khoshgoftaar, T. M., & Van Hulse, J. (2009). A study on the relationships of classifier performance metrics. Paper presented at the 2009 21st IEEE international conference on tools with artificial intelligence.
[41]Sherstinsky, A. J. P. D. N. P. (2020). Fundamentals of recurrent neural network (RNN) and long short-term memory (LSTM) network. 404, 132306.
[42]Staudemeyer, R. C., & Morris, E. R. J. a. p. a. (2019). Understanding LSTM--a tutorial into long short-term memory recurrent neural networks.
[43]Tsoumakas, G., Katakis, I., & Vlahavas, I. (2006). A review of multi-label classification methods. Paper presented at the Proceedings of the 2nd ADBIS workshop on data mining and knowledge discovery (ADMKD 2006).
[44] Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., . . . Polosukhin, I. J. A. i. n. i. p. s. (2017). Attention is all you need. 30.
[45]Wang, Y., Gao, J., & Chen, J. (2020). Deep learning algorithm for judicial judgment prediction based on BERT. Paper presented at the 2020 5th International Conference on Computing, Communication and Security (ICCCS).
[46]Yang, M. (2020). ckiplab/albert-base-chinese. from https://huggingface.co/ckiplab/albert-base-chinese
[47]Yang, M. (2020). ckiplab/bert-base-chinese. from https://huggingface.co/ckiplab/bert-base-chinese