摘要
目前已經有許多研究嘗試讓電腦像人類一樣說故事，但是對於故事本身仍然沒有一個完整且結構化的定義。敘事學家建議一個故事應該分為三層：故事素材層、情節層和表現層，而Swartjes為故事素材層定義了初步的框架。然而，這個框架只定義了六種基本元素和四種因果關係，仍然缺少許多說故事時必要的資訊，例如元素間的時間關係。 在本文中，我們在原本的故事素材模型上定義了元素間的時間關係，並且提出一組規則，用來判斷在什麼條件下應該在兩個元素之間加入時間關係。最後，我們利用一種類似深度優先搜尋法（DFS）的演算法將基於原本模型的故事素材和加入時間關係的故事素材轉成人類能閱讀的文章，並讓一般觀眾來評判那個故事素材產生的故事更容易理解。得到的結果是：從加入時間關係的故事素材產生的文章都得到較高的評價，而且對這些文章的意見也較為集中。
並列摘要
Nowadays, there are many researches trying to make computers tell stories. But there is no complete and structural definition for stories. The narrative theorists suggested that a story should be divided into three layers: fabula, lot, and presentation. And Swartjes offered an initial framework for the fabula layer. However, this framework only define six basic elements and four causal relations and still lack a lot of information that is necessary for storytelling, like the temporal relation between two elements. In this thesis, we define the temporal relation between two elements on original fabula model and construct a set of rules, which are used to determine under what condition we should join the temporal relation between two elements. Finally, we use an algorithm which is similar to Depth-first search (DFS) to transform fabula instance based on original model and instance with temporal relation into human-readable articles. And we let general viewers judge which article is more understandable. The results are that the articles from fabula instance with temporal relations get higher scores, and the opinions to these articles are more centralized.