{"course":{"id":"115162701X01","name":"自然語言處理","enroll_type":"選修","general_type":"","dept":"人工智慧學程","grade":4,"class_group":"甲","teacher":"林淯慈","credit":3,"hour":3,"time":[{"day":5,"section":[2,3,4]}],"classroom":"","max":25,"min":10,"remark":"","alt_for":"","courses":""},"detail":[["指定教材","1. 教師自編講義\n2. Hugging Face 官方技術文件與開源模型手冊 (Hugging Face Official Documentation & Model Guides)"],["參考教材","1. PyTorch 自然語言處理|以深度學習建立語言應用程式 Delip Rao, Brian McMahan 著 ()，楊尊一 譯。2019年6月版Gotop 碁峰資訊。ISBN 9781491978238 \n2. PyTorch 官方使用者手冊與 API 文件 (PyTorch Official Documentation)\n3. scikit-learn 官方機器學習手冊 (scikit-learn User Guide)"],["成績評量標準","1. 平時實作與 GitHub Commit 紀錄  30%\n2. 期中評量 30% [課堂測驗 (15%) 資料前處理報告DATA_REPORT.md (15%)]\n3. 期末專案與發表 40% \n    -  專案發表 15%\n    -  AI 協作日記 AI_USAGE.md 15%\n    - 最終專案履歷 10%"],["教學方法","理論講授,實務講授,個人實驗(實作)"],["教學目標","1. 建立自然語言處理之核心理論基礎。理解自然語言處理之演進與核心概念，從傳統詞向量、機器學習分類器（scikit-learn），到現代深度學習序列模型（PyTorch RNN/LSTM）與 Transformer 架構。\n2. 掌握現代 AI 輔助軟體工程工作流程。引導學生使用 VS Code、GitHub 與 AI 程式碼助手進行「Vibe Coding」。學員將具備運用生成式 AI 輔助開發、除錯、版本控制及自動化建置 Python 專案的實務工程能力。\n3. 實作檢索增強生成與系統整合應用。\n4. 鏈結 iPAS 認證與 AI 系統規劃治理能力。訓練學生在開發過程中，考量模型幻覺風險、API 成本預估（Token Economics）、資訊安全防護及輕量化微調（PEFT/LoRA）之系統部署效益。"]],"office_time":[["林淯慈","Friday 4 - 6 pm"]],"schedule":[["2026/09/11","2","4","3","林淯慈","自然語言處理導論與開發環境建置。Introduction to Applied NLP & Development Environments","GitHub 基礎專案架構  與 README.md","IDE 配置、PyTorch 環境建置與 Git 版本控制工作流程。AI 輔助軟體工程方法論與反覆運算流程。 Local IDE configuration, PyTorch CPU/CUDA environments, and Git version control workflows.",""],["2026/09/18","2","4","3","林淯慈","文字處理基礎與非結構化資料獲取。 Fundamentals of Text Processing & Data Acquisition","01_data_extraction.ipynb 與原始語料整合。","企業非結構化與結構化語料之本質探討。 • PDF 文件之演算法解析與文本擷取。文件格式異常與擷取雜訊之評估。",""],["2026/09/25","2","4","3","林淯慈","國定假日 / 放假","","",""],["2026/10/02","2","4","3","林淯慈","文字正規化與 PyTorch 張量表示法。Text Normalization & Tensor Representation in PyTorch","02_text_normalization.ipynb 與張量化標準語料集","分詞、詞性標註與命名實體識別基礎。 詞向量架構（Word2Vec）與將詞元序列轉換為 PyTorch 張量（torch.Tensor）。",""],["2026/10/09","2","4","3","林淯慈","國定假日 / 放假","","",""],["2026/10/16","2","4","3","林淯慈","鑑別式機器學習與基礎神經網路模型","03_classification_baseline.ipynb 與分類評估指標報告。","監督式機器學習演算法：邏輯迴歸、決策樹與隨機森林。 多層感知機 (MLP) 與人工神經網路 (ANN) 原理","使用 scikit-learn 與 基礎 PyTorch 多層感知機建立評估基準線。"],["2026/10/23","2","4","3","林淯慈","神經語言模型與 PyTorch 序列模組。","04_attention_heuristics.ipynb","深度學習在語言模型之應用：RNN 與 LSTM 架構。 為解決序列瓶頸而引入之「注意力機制 (Attention)」概念。。","使用 PyTorch 實作"],["2026/10/30","2","4","3","林淯慈","Transformer 架構與生成式 AI 系統","05_zero_shot_inference.ipynb，呈現大型語言模型與基準模型之比較。","Transformer 編碼器（BERT）與解碼器（GPT）架構解析。 自注意力機制 (Self-Attention)、多頭注意力與 Q-K-V 映射原理。透過 Hugging Face 與 LLM API 進行預訓練模型推論。",""],["2026/11/06","2","4","3","林淯慈","期中評量","1. 資料前處理報告 DATA_REPORT.md (Midterm Data Preparation Report in git repository).     2. 課堂測驗。in class test.","DATA_REPORT.md: 領域選擇、清理對比截圖與資料規模","版本控制：建立 Git 標籤與發布 GitHub 版本。 (Versioning: Creating Git tags and publishing GitHub releases)"],["2026/11/13","2","4","3","林淯慈","密集向量嵌入與向量化資料結構。 Dense Embeddings & Vectorized Data Structures.","06_vector_similarity.ipynb","從詞嵌入到句子級密集向量之技術演進。 高維度向量空間塑模與 cosine 相似度計算。","Cosine similarity metrics and PyTorch sentence-transformers.  建置向量化資料結構與本機向量索引庫。"],["2026/11/20","2","4","3","林淯慈","大語言模型與知識擴充技術（RAG：文本切分與索引）","07_vector_indexing.ipynb","上下文視窗限制與檢索增強生成（RAG）架構原理。 長篇行政法規之文本切分（Chunking）演算法與重疊策略。 建置本機向量索引資料庫以供檢索管道調用。",""],["2026/11/27","2","4","3","林淯慈","情境學習與端到端 RAG 系統整合","08_rag_pipeline.ipynb","預訓練模型推論典範與 In-Context Learning技術。 動態檢索上下文注入、結構化提示詞設計與原始佐證資料引用。 模型幻覺與非事實輸出的演算法防護機制",""],["2026/12/04","2","4","3","林淯慈","系統快速Prototyping & Applied Implementation","app/app.py & screenshots/demo_ui.png","將後端 PyTorch/LLM 運算流程封裝為互動式 Web 應用原型。 使用 Streamlit 快速構建低程式碼對話式使用者介面。 對話歷史狀態管理與即時串流推論顯示。",""],["2026/12/11","2","4","3","林淯慈","參數高效微調技術與 AI 治理實務。","09_lora_analysis.ipynb and governance documentation in README.md.","全參數微調與基於 PyTorch peft 套件之 LoRA 輕量微調技術 (Low-Rank Matrix Decomposition)。 AI 治理與安全：越獄攻擊、提示詞注入與資料隱私防護實務。",""],["2026/12/18","2","4","3","林淯慈","期末評量：實作成果發表。","建立 Git 發布標籤 v1.0-final，包含經完整驗證之程式碼與 AI_USAGE.md。","實作成果展示與技術挑戰答辯。",""],["2026/12/25","2","4","3","林淯慈","國定假日 / 放假","","",""],["2027/01/01","2","4","3","林淯慈","國定假日 / 放假","","",""],["2027/01/08","2","4","3","林淯慈","自主學習: iPAS AI 應用規劃師專業認證準備與複習。 Asynchronous Self-Study for Professional Certification Preparation.","最終 GitHub 專案履歷 (Portfolio) 連結。","iPAS AI 應用規劃師核心考核能力指標整合複習。",""]]}