申請模型存取權限 Request for Model Access
您好：
非常感謝您投入心力開發並開源 privacy-filter-tw。我們近期正進行企業地端大型語言模型（Local LLM）的測試、導入與應用建置，而繁體中文及台灣在地個資格式的去識別化，是整體資料治理與隱私保護流程中相當關鍵的一環。
經過初步評估後，我們對貴專案針對台灣情境所提供的 PII 偵測能力，以及模型在實務上的應用潛力深感肯定。後續預計將此模型應用於公司內部資料前處理，例如 RAG 知識庫建置、文件檢索及內部資料分析等情境。所有資料皆會先於地端完成個資偵測與自動遮罩，再交由地端 AI 系統進行後續檢索、分析或生成，以降低敏感資訊暴露風險並強化隱私保護。
本模型僅規劃用於內部測試及企業內部應用，不會將原始資料或模型處理結果對外公開。若有任何使用規範、授權條件或建議的實作方式，我們也願意配合遵循。
懇請您協助審核並核准存取權限。再次誠摯感謝您的開源貢獻，以及您對繁體中文、台灣在地 PII 去識別化技術的投入與推進。
敬祝
順心平安
Hello,
Thank you very much for your work in developing and open-sourcing privacy-filter-tw. We are currently evaluating, deploying, and building enterprise on-premises Large Language Model (Local LLM) applications. De-identification of Traditional Chinese text and Taiwan-specific personal data formats is a critical part of our data governance and privacy-protection workflow.
Following our initial evaluation, we have been highly impressed by the project’s PII detection capabilities for Taiwan-specific use cases and its practical potential for real-world deployment. We plan to use the model for internal data preprocessing, including RAG knowledge-base construction, document retrieval, and internal data analysis. All data will first undergo local PII detection and automatic masking before being processed by our on-premises AI systems for retrieval, analysis, or generation. This is intended to reduce the risk of sensitive-data exposure and strengthen privacy protection.
The model will be used solely for internal testing and enterprise-internal applications. Neither the original data nor the model outputs will be disclosed publicly. We are also willing to comply with any applicable usage guidelines, licensing requirements, or recommended implementation practices.
We would greatly appreciate your review and approval of our access request. Thank you again for your open-source contribution and for advancing PII de-identification technology for Traditional Chinese and Taiwan-specific contexts.
Sincerely,
Ray