Ontology可用來組織、管理與分享知識，Ontology Engineering是一種建構Ontology的過程，建構的過程中，多數的工作需要人費時費力地去完成，因此利用機器來輔助Ontology Engineering成了一門重要的課題。使用Knowledge Discovery的方法協助Ontology Engineering建構Ontology的過程，稱為Ontology Learning，本論文中提出的Ontology Learning方法為分析使用者在搜尋引擎下關鍵字查詢時的行為，加上利用與查詢關鍵字有關的網頁資訊，以輔助建構Ontology。本論文中的Ontology由使用者所查詢的關鍵字組成，我們要learning的，則是這些關鍵字彼此之間的關係，其中有上義詞、下義詞與同義詞等等，因此，自動尋找關鍵字彼此之間的關係以輔助建構Ontology，即為我們提出本論文的目的。除此之外，本論文亦實作了完整的Ontology Learning系統，從一開始使用者查詢記錄的蒐集，關鍵字擷取與分析，關鍵字之間的關係判定，直到最後Ontology的產生，都將由系統自動完成。 / Ontology can be used to organize, manage and share knowledge. Ontology Engineering is the process of constructing Ontology. However, it’s usually a time-consuming and error-prone task. Thus, utilizing methods of Knowledge Discovery to help Ontology Engineering is called Ontology Learning. In this thesis, Ontology Learning process is done by using those pages related query terms and analyzing the querying behavior of users on search engines. The Ontology is organized by user query terms and relations among them. These relations we define are hyperonomy, hyponomy, synonymy and et al. Our goal of this thesis is to automatically learn the correct relations among these query terms. Besides, we implemented the complete system platform for Ontology Learning. The system can automatically collect logs, extract and analyze query keywords, and produce the final Ontology.

搜尋引擎已成為大多數人普遍使用的電子服務項目。Google是全球網路搜尋的龍頭，但是Google在中國、南韓和俄羅斯三個國家未竟全功，而分別由百度、Naver和 Yandex三個搜尋引擎獨占鰲頭，有著相同語言和文化知識的本國搜尋引擎，成為當地使用網路者之首選。Yandex是俄羅斯排行第一搜尋引擎，尤其是使用俄文搜尋時；也是俄羅斯最具創新的公司。本研究藉由分析Yandex本身及進一步比較如百度和Naver的本土搜尋引擎，探討Yandex成為俄羅斯網路搜尋龍頭的最主要因素。 百度、Naver和Yandex這三間公司會成為當地最重要的搜尋引擎，其原因最大的共通性是「高度本地化的服務」。相較於Google這類外國公司，更了解當地市場的需求。以Yandex為例，其擁有突破性的技術，專注於搜集且貼近當地需求，充分發揮在地化的服務，進而保持在市場上的領導地位。 Yandex未來發展將專注於擴大搜尋服務到新興市場，精進其於移動應用服務及因應全球趨勢。身為一個在俄羅斯網路搜尋界的領導者，沒有其他競爭者比Yandex更會培訓旗下的企業者，運用機會，成功地吸引到本地網路使用者。 / Search engines have become usual service for the large majority of people. Google is acknowledged leader in global Internet search. There are three countries where Google has failed to dominate the search market. Baidu, Naver and Yandex are clear market leaders in China, South Korea and Russia respectively, where local language and cultural knowledge has made them first choice for many web users in those markets.Yandex is a Russian based search engine focused on search in Russian language and number one innovative company in Russia.Through business analysis of Yandex itself, and other local search engines Baidu and Naver, and their further comparison, this research discusses the main factors that made Yandex a leader in Russian Internet search. Through the case study the author concluded the factors led to dominance on local markets of each search engines discussed have similarity among three companies Yandex, Naver and Baidu. Their highly localized services brought them comparatively stable and long-term market leadership. All of those companies know local markets better than foreign based competitors such as Google. As for Yandex, breakthrough technology, focus on search and better market understanding and localizing all the services played sufficient role in keeping a leadership position on the market. This study suggested Yandex future development should be focused on expanding into new geographical markets in search industry, as well as will be highlighted with expanding its mobile applications service along with global mobilization trend. As a leader in Russian Internet search, there is nobody better placed than Yandex to educate business practitioners in the Internet on how to tap into those opportunities and successfully target the local audience.

「民以食為天」，飲食與我們的生活息息相關。近年來由於食安風暴肆虐，自行烹煮的需求隨之高漲。然而在家自行烹煮時常會面臨不知道該烹煮什麼料理的問題，因此有便利的食譜搜尋系統對烹煮的人而言將是相當方便的。然而使用搜尋系統時，由於我們只知道想用某些特定食材進行烹煮，而不知道哪些食譜含有特定食材，因此在以少數食材進行查詢時不免會得到過多的食譜結果而難以快速找到喜好的食譜。我們建立了一個食譜搜尋的自動完成機制，並依照該機制實做出了食譜搜尋引擎。使用者使用系統進行搜尋時，我們將會依照使用者輸入的食材尋找適合搭配的食材推薦給使用者，幫助使用者在查詢時使用更完整的Query讓搜尋系統可以找到更少更精準的食譜，幫助使用者更快的找到喜歡的食譜。然而只推薦搭配性食材，可能會推薦出與Query中的食材是替代關係的食材，也就是通常不會一起出現的食材，因此我們也進行了替代性食材的研究。給定由兩個食材組成的食材配對，我們研究如何自動的判斷替代性食材。我們將問題轉化成分類問題來解決，並使用One-Class Classification的技術解決分類問題中的Imbalanced Problem。我們使用f1-score觀看One-Class Classification與傳統分類器的比較。經實驗測試，One Class Classification與傳統分類器相比，One Class Classification較能協助我們解決Imbalanced Problem。

近年來，動漫(動畫、漫畫與線上遊戲)越來越受歡迎。全球資訊網也陸續出現收集大量包括故事情節、動漫角色、作者等動漫相關資訊的動漫網站。多數動漫網站都提供使用者文字檢索的功能，以搜尋動漫網站文字內容。但是動站網站若能提供根據文字與圖形風格來搜尋圖文內容，對於動漫使用者而言，將更為方便。圖文風格可能是漫畫人物的繪畫風格、動畫故事的敘事風格等等。 為了方便使用者以圖文風格進行搜尋相關資訊，本論文根據動漫關連探勘技術，研究並開發一個動漫網站的搜尋引擎：動漫魔鏡，以提供使用者根據圖文關連來搜尋動漫網站中風格相似的動漫資訊。本論文的搜索方法關聯了圖像特徵和文字特徵。首先，針對圖像特徵，由於動漫角色是動漫的靈魂，因此經過動漫臉部偵測後，我們以電腦視覺中的局部二值樣式(Local Binary Pattern, LBP)與灰階值分佈來抽取並表示動漫角色的臉部特徵。針對文字特徵，我們利用一般全文檢索技術來擷取文字特徵。接著，運用階層式分群技術將文字特徵與圖像特徵值轉換為文字與圖像關鍵詞。最後，以語意主題模型中的隱含狄利克雷分布(Latent Dirichlet Allocation, LDA)分析圖文關鍵詞的潛在語意，並據此計算動漫網頁之間的相似性。實驗結果顯示本論文所研發的風格搜尋，其效果優於其他三種基本作法。 / Animations, comics, and games (ACG) have become more and more popular in recent years. There exist many ACG web sites which contain lots of textual and visual information on stories, characters and authors of animations, comics and games. Most ACG web sites provide users text retrieval capability to search for textual contents. However, there is a need for users to search for textual and visual contents by styles. Examples of styles are drawing styles of comic characters, narrative styles of animation stories and so on. In order to help users to search for textual and visual contents by similar styles, this thesis investigates and develops a search engine, Comirror, for ACG web sites based on latent correlation between textual and visual contents. First, while facial styles of characters play important roles in ACG, after comic face detection, Local Binary Pattern (LBP) along with gray-value histogram is utilized to extract and represent the visual features. For the textual contents, traditional full-text indexing technique is employed to extract textual features. Then, hierarchical clustering is performed to quantize and transform the textual and visual features into textual and visual words. Finally, Latent Dirichlet Allocation (LDA) is utilized to discover the latent semantic correlation between visual and textual words. Experiments show that the developed approach performs better than the other baseline approaches.

網際網路在許多方面皆改變了大眾的讀報方式：不僅僅是因為Google News一類的新聞搜尋引擎，提供讀者多樣化且一次性的即時整合型閱覽，現今多數的網路使用者，也已養成從數個不同網頁中選擇所需資訊來瀏覽或進行研究、調查的習慣。然而，隨著報紙印刷發行量的銳減、新聞網站的流量不如預期，新聞媒體出版商對於搜尋引擎憑藉他人著作獲取暴利的手法感到越來越挫敗。Google自西元2009年起在Google News美國版刊登廣告，會不會成為壓垮傳統新聞媒體的最後一根稻草?而Google的這項舉措是否動搖其利用他人著作內容賺錢的正當性?可符合美國著作權法的合理使用原則? 本文將就Google News刊登廣告所引發的爭議，從網路新聞的發展、Google的經營模式以及合理使用原則的規範目的等面向，鎖定美國著作權法，來探討網路搜尋引擎截取他人著作內容的合法性，以及數位時代下合理使用原則的未來命運。 / The Internet has changed the way people reading the newspaper: not only because news search engines such as Google News providing readers with a diverse and one-time real-time integration of reading, but also because nowadays, most Web users have become accustomed to selecting the information they need from several different Web pages for browsing or carrying out research, and investigation. However, as the newspapers’ circulation drops, and the flow of news Web sites does not reach the expectation, the fact that search engines use others’ works to obtain profits disappoints news media publishers a lot. Will the advertisements placed alongside search results on the American version of Google News be the last straw? Will this undermine the validity of Google’s initiative of using others’ work for money, and fit in with the fair use doctrine? This article will try to analyze the Google News advertising disputes from the perspectives of development of network newsm, Google's business model and the fair use doctrine , and to discuss the legality of the Web search engine as well as the fate of fair use under digital era in the future.

源於資訊量爆炸時代的來臨，企業面臨大量資料所帶來的挑戰：傳統關聯式資料庫無法負荷龐大資料所造成的效能及儲存設備升級等問題。為了解決大量資料所帶來的諸多問題，各界提出不同的理論，而其中最被廣為討論的就是雲端運算。時至今日，許多企業及個體用戶逐漸開始使用雲端運算中，目前最具代表性的分散式架構Hadoop上的資料庫代表－欄導向資料庫HBase來作為底層資料庫。故本研究提出一套傳統關聯式資料庫轉換至欄導向資料庫HBase之轉換機制，以台灣學術期刊搜尋引擎為例。

本研究以2001年至2013年諾貝爾物理學獎得主之著作為研究樣本，比較八個商業資料庫(Scopus和Web of Science)與開放取用系統(搜尋引擎：Google Scholar、Microsoft Academic；匯集式機構典藏系統：OpenDOAR、OAIster；學科性開放取用系統：arXiv.org和Astrophysics Data System)，於物理學文獻收錄之正確性、完整性、重複性(包含內部重複與外部重複性)和獨特性，並評析各資料庫與系統之檢索功能、資料呈現等面向。期望能對圖書館資料庫選購以及使用者檢索資料庫與系統提供建議，並為各資料庫與系統之未來發展提出建議。 研究結果顯示：(一)諾貝爾物理學獎得主之個人著作揭露情形尚未普遍；(二)商業資料庫檢索功能較為多元，搜尋引擎容錯機制較強；(三)開放取用系統Astrophysics Data System和Microsoft Academic改版上線後，檢索功能Google化，重視全文鏈結、圖像化資訊呈現與語意網連結資訊；(四)各資料庫與系統普遍出現書目著錄格式不統一之問題，影響書目品質與檢索效率；(五)一般而言搜尋引擎資料完整性高於商業資料庫，商業資料庫高於機構典藏系統，但學科性開放取用系統Astrophysics Data System之資料收錄完整性僅低於Google Scholar；(六) arXiv內部重複性最低，Google Scholar和OpenDOAR內部重複性最高；(七)開放取用系統彼此重複性高，且與搜尋引擎Google Scholar和Astrophysics Data System重複性達100%。由於各資料庫與系統之收錄範圍各不相同，不同資料庫與系統亦提供不同的功能，使用者應依個人資訊需求與目的選擇資料庫與系統使用，如欲檢索物理學文獻，使用搜尋引擎與開放取用系統Astrophysics Data System可獲得較完整之文獻：若使用者欲取得引文分析之相關資訊，則以選擇商業資料庫Scopus和Web of Science為佳，亦可選擇Astrophysics Data System。 / In this study, scholarly communication system of commercial services and open access will be examined through comprehensiveness, overlap and database variation of coverage via field operations of commercial citation index databases (Web of Science and Scopus) and open access citation system (search engine: Google Scholar and Microsoft Academic; disciplinary of physics: arXiv.org and Astrophysics Data System; prestigious institutional repository: OAIster and OpenDOAR). Retrievals will be conducted in the two commercial databases, two search engines, and four open access systems stated above to analyze and compare their retrieval interfaces, and evaluations of each system will be made as well according to presentation and output of retrieval results. Noble laureates in physics sciences from 2001 to 2013 are selected as samples in this study. Records of their publications over time will be retrieved and downloaded from each system, and a computer program will be developed to perform the analytical tasks of sorting, comparison, elimination, aggregation and statistics. Bibliographic records retrieved from the two databases and six systems will undertake quantitative analyses and cross references to determine the comprehensiveness and uniqueness of their system coverage. The results of the study may provide better references for libraries to acquire citation index databases, to build institutional repositories, or to create citation index systems on their own in the future. Suggestions on indices and tools for academic assessment will be presented based on the comprehensiveness assessment of each system as well.

近幾年來，有關地理資訊的網頁搜尋越來越受到重視。傳統的網頁搜尋引擎無法反應使用者查詢和網頁文件之間的地理關聯性。在一些情況下，我們希望網路搜尋引擎能夠考慮使用者查詢與網頁文件間的地理相關性，以提升搜尋的準確度。 我們的研究透過包圍矩形模型(Bounding Rectangle Model；BR Model)以搜尋與使用者查詢之地理相關程度較高的網頁文件。 使用者僅需輪入文字的查詢，即能得到相符結果的網頁文件。首先，我們建立一個地名辭典以找出使用者查詢與網頁文件內出現的地名及空間資料，接著我們利用空間資料建立空間索引項(spatial index term)集合，用來表示使用者查詢與網頁文件內的地理範圍，最後再透過使用者查詢與網頁文件的空間索引項集合計算兩者之間的地理相似程度，以找出與使用者查詢有較高地理關聯性的網頁文件。 此篇論文的貢獻在於我們提一套完整資訊檢索模型架構的方法，分析使用者查詢與網頁文件之間的地理關聯性，使用者藉由輸入文字查詢即能得到相符地理關聯性的網頁文件。 / Geographic web search becomes increasingly popular in recent years. Traditional web search engine, such as Google and Yahoo, can not accommodate geographic relevance between user queries and internet documents. Hence, they can not retrieve geographic related information from user queries. However, in many cases, the geographic relevance between user queries and internet documents could enhance the accuracy of this type of searches. In this thesis, we propose a mechanism that uses the Bounding Rectangle Model (BR Model) to retrieve geographic relevant internet documents in response to user queries. Users provide only the conventional input queries (keywords) and our search engine will return the geographic relevant results. Our method can be classified into the following three steps. In the first step, we create a gazetteer and use it to relate the user query’s geographic terms in internet documents. In the next step, we use the spatial data to build a set of spatial index terms that represents the geographic scope of user query and internet documents. And then we use these spatial index terms to calculate degree of geographic similarity between user query and internet documents to identify highly relevant geographic internet documents. We implemented a prototype search engine using our approach. The experiment results show that we can successfully retrieve geographic relevant data through this mechanism and provide more accurate search results.

在網路的世界裡，搜尋引擎對網路使用者的重要性逐漸升高，而藉由搜尋引擎之幫助，使用者可以在數位資訊爆炸的網路世界中，尋找到所需的資訊。惟此種便利之用途，亦造成著作權人之權利遭受侵權之疑慮，如搜尋引擎協助使用者找到未經合法授權之檔案而下載之；亦或搜尋引擎本身即複製該未經合法授權之內容而散布之等情形。各國對於搜尋引擎業者之侵權責任有相似規定，在如何之情況下搜尋引擎必須負擔侵權行為責任，而在何種情形下搜尋引擎業者可能可以免責，又在何種情形下搜尋引擎業者可能引用合理使用原則，本文亦嘗試提出各國之規定以供國內參考。 國內尚未有搜尋引擎業者之相關案例，故本文嘗試透過不同的國家中不同的搜尋引擎態樣進行案例分析，提出不同的搜尋引擎在不同之情狀之下，可能會造成的著作權爭議。如在圖像搜尋方面即提出美國的Perfect 10 v. Google案中縮小圖示之合理使用爭議，並同時提出多年前的Kelly v. Arriba案做為對照；中國大陸的七大唱片公司訴百度案在音樂搜尋當中是相當具有爭議的問題，本文亦提出其他相關的大陸案例做為對照；瑞典海盜灣的案件是P2P軟體變型後所產生之搜尋引擎，亦可以看出未來會有更多不同態樣的搜尋引擎出現；而新聞搜尋之問題，牽涉到重製權，近期Google與美聯社達成和解，本文亦為新聞搜尋此特殊的態樣為探討；而圖書搜尋方面則是以Google的圖書館計畫為基礎，並進一步探討合理使用的可能性。 本論文試圖提出美國、歐盟以及中國大陸的法規以及諸多案例，並探討在世界各地的搜尋引擎面對的著作權爭議問題，而這樣的問題會隨著科技的進步與發展有更多不同態樣的變化。我國目前尚未有相關的案例，故本文嘗試提出案例語法規的論述與探討，以期能予我國參考之。 / In the internet world, search engines become much more important to the users. By using the search engines, users can easily gather the useful data online. Although search engines bring so much convenience, but the copyright holders accuse the search engines of coypright infringement. For instance, the search engines sometimes help their users to download the copyrighted materials or themselves reproduce the copyrighted materials. The actions might bring them law suits.There are similar regulations of search engines in different countries. In some situation, search engines might defense themselves by claiming safe harbor regulation, and in others, they could claim the fair use doctrine. In the thesis, it states many cases from different countries to discuss the regulations. Due to the different functions of search engines, there are various cases. Kelly v. Arriba case and Perfect 10 v. Google case are about thumbnails search in the United States. Baidu v. EMI record case in China is about mp3 files search. Despite the traditional search engines, there are some relatively new and specific search engines, such as Private Bay. Private Bay is a search engine which focuses on searching bittorrent files for P2P downloading. In the future, there will be much more search engines with different functions. In the thesis, it also states the disputes of news search and book search.Google Book Project rised discussion worldwide. The thesis states the opinions of the ones who agree to apply the fair use doctrine on the project and the one who disagree to. In Taiwan, there have not been any copyright cases of search engines so far.The thesis states the copyright regulations in U.S., Europe and China, such as safe harbor, contributory infringement and fair use doctrine. By presenting the cases and regulations in other countries, the thesis also provide some inspirational ideas about the copyright disputes of search engine.

西元2010年Google不滿中國對於網際網路之審查規範，因而宣布退出中國網際網路搜尋引擎市場之營運，Google認為中國網際網路之審查規範已影響自由貿易，並於同年（2010年）發表白皮書，強調政府管制網路資訊的行為會影響貿易，於附件中討論WTO服務貿易總協定應如何適用於政府限制網路資訊之措施。Google冀望美國能將此爭議訴諸於世界貿易組織之爭端解決機構；而美國貿易代表署也受Google之呼籲，著手研究將中國網路審查措施，向世界貿易組織提出爭端案件之可能。 為了解本爭端若訴諸WTO爭端解決，美國勝訴之可能性，以及預測中國是否可能因Google案件而被迫調整其網路管制以符合WTO規範，本文討論重心將從WTO服務貿易總協定規範之角度探討Google所指摘之中國對於網際網路搜尋引擎服務之管制。