摘要 Abstract 
近年來機器學習時常用於風格傳播，而傳統人工設計一套字型，從初始的設計字型就需要專精的技術， 且後期需要耗費大量的人力和工時，來保持字與字之間統一特定的風格，因此普通人設計一套完整的漢字字體庫也成為工程浩大的任務。
In recent years, machine learning is often used for style transfer, while the traditional manual design of a set of fonts requires specialized skills from the initial design of the fonts, and requires a lot of labor and man-hours to maintain a uniform and specific style among the font.
考慮到漢字越來越普及，為了解決設計漢字過程繁瑣的難題，本論文以內含漢字的字體庫為出發點， 開發兩套基於機器學習的字型SVG生成系統，其生產出的SVG可打包成字型檔做使用，不但確保生成圖像的風格統一性，更減少了人力資源，並且大幅地提高產出、調整的效率。
Considering the increasing popularity of Chinese characters, in order to solve the tedious problem of designing Chinese characters, this paper takes the font containing Chinese characters as the starting point and develops two sets of machine-learning-based font SVG generation systems, whose generated SVG can be packaged into computer font, not only to ensure the uniformity of the generated images, but also to reduce human resources and significantly improve the efficiency of production and adjustment.
此外，除一般常見的黑白字型檔之外，本論文也針對Adobe近年所推出的OpenType-SVG，提供了保留顏色資訊的字型檔，在Illustrator、Photoshop等特定平台上可以顯示有別於傳統字體的彩色效果，透過色彩的轉變展現字體設計的多樣性。
In addition to the common black and white font files, this paper also focuses on the OpenType-SVG introduced by Adobe, providing font files that retain color information and can display color effects different from traditional fonts on specific platforms such as Illustrator and Photoshop. Through the change of colors, we can show the diversity of font design.
