隨著網路與社群媒體的興起，線上百科已是現代人日常生活重要之知識來源。維基百科因其自由開放與共享之精神，是目前網路上最大且最重要的百科全書知識庫。然中文維基百科雖然參與者眾，其條目數與西方主流語言的維基百科相比，仍然存在數倍以上的巨大差距，幾乎不可能以人工弭平。本計畫期望能以人工智慧的方法弭平這種知識差距。我們的構想是：研發人工智慧的技術，根據其他多種語言的條目，自動生成中文條目及內容，以擴充中文維基百科的知識涵蓋率，從而實現中文知識內容的流通與傳播。我們的發展步驟如下：第一年，我們將發展跨語言英中維基百科條目內容改寫技術，透過「跨語言階層式 Transformer」，將英文內容轉換生成為中文條目內容。第二年，我們擬發展基於循環生成對抗網路之文章風格轉換方法，自動學習如何依據目標用戶的習慣調整風格。第三年，我們將發展利用多語條目及外部資料之自動改寫技術。此技術利用維基百科多語種之優勢，將不同語言版本的百科知識以中文呈現，進而使中文維基百科的內容更加完備。 ;With the rise of Internet and social media, online encyclopedia has become an important source of knowledge in modern people's daily life. Wikipedia is the largest and most important encyclopedia knowledge base on the Internet because of its free and open spirit. Although Chinese Wikipedia has a large number of participants, its number of entries is still far less than the number of Wikipedia entries in mainstream Western languages (such as English). This research project is expected to develop methods for automatically generating Chinese encyclopedia entries based on other language entries to expand the Chinese Wikipedia knowledge base, thereby achieving the circulation and dissemination of Chinese knowledge content. Our plan is as follows. In the first year, we will develop cross-language English-Chinese Wikipedia entry content rewriting technology and use the "cross-language hierarchical Transformer" to convert English content into Chinese entry content. In the second year, we plan to develop an article style conversion method based on recurring adversarial networks, and automatically learn how to adjust the style according to the taste of target users. In the third year, we will develop automatic rewriting technology that uses multilingual entries and external sources. This technology takes advantage of Wikipedia's multilingualism and presents encyclopedic knowledge in different language versions in Chinese, thereby making the content of Chinese Wikipedia more complete.
