Automatic Workflow Base on Web Knowledge Extraction - A Case Study on Recipe Website

[[abstract]]健康是樂活主義中極為重要的一環。在這股風潮下，市面上出現了不少食譜書籍與網站介紹健康料理，但由於現代人大都過著忙碌的生活，又此類食譜書與網站大多只提供單一料理建議，亦沒有提供使用者一個完整性的料理規劃。例如，把要使用的食譜，將所有的步驟與材料做有效率的規劃，讓使用者節省搜尋與整理的時間。透過資料探勘的技術來分析網站資訊，建立知識地圖，融合成為較完整的領域知識本體論 (Domain Ontology, DO)。除了提供領域知識建構外，有可以根據Ontology提供使用者更友善的瀏覽動線，能夠更迅速且正確的就找到他們想要的結果。基於這些動機，本論文利用食譜網站做為應用領域，整合領域內容和知識的系統，首先，藉由搜尋分析的方法，有效的將該領域的重要有效的將抓取領域的重要網站的所有網頁，並分析HTML網頁內容分析主要內容區塊，快速的擷取與整合該領域網站的metadata，即能準確擷取食譜metadata。探索概念間之關聯，建立食譜內容的初步Ontology，並擴充成為食譜的知識或學習地圖。藉此，可以提供讀者快速概觀食譜。以SVG (Scalable Vector Graphics) 為基礎的圖形化知識地圖介面，最後，本論文也將metadata擷取成果，有效應用於建立DO內概念的關聯性，提供自動建立食譜領域知識的系統雛形。[[abstract]]This thesis focuses on the recipe application domain, in which the proposed system collects recipe-related pages from websites, extracts domain metadata from pages, mines relations among recipe data objects, and integrates those information as the knowledge for recipe domain. First, by searching the Web and analyzing pages and sites, the system accumulates numerous recipe data. Then, major content blocks of recipe pages are extracted by analyzing the HTML DOM tree structures for pages and sites. By building indices for texts extracted from those pages, phrase extraction and paring methods are employed to identify significant recipe keywords (key phrases) as domain concepts. Mining associations between recipe-domain concepts is applied to explore concept relations to build the domain ontology. Recipe concepts and relations are presented as the Knowledge Map using SVG (Scalable Vector Graphics). Finally, the recipe domain knowledge is applied to provide services for mobile applications as the demonstration..[[note]]碩