影像中的文字訊息，通常包含著與場景內容相關的重要資訊，如地點、名稱、指示、警告等，因此如何有效地在影像中擷取文字區塊，進而解讀其意義，成為近來電腦視覺領域中相當受矚目的議題。然而在眾多的場景文字偵測方法裡，絕大多數是以英文為偵測目標語言，中文方面的研究相當稀少，而且辨識率遠不及英文。因此，本論文提出以中文和英文為偵測目標語言的方法，分成以下四個主要程序：一、前處理，利用雙邊濾波器(Bilateral filter)使文字區域更加穩定；二、候選文字資訊擷取，考慮文字特徵，選用Canny 邊緣偵測和最大穩定極值區域(Maximally Stable Extremal Region)，分別提取文字邊緣和區域特徵，並結合兩者來優化擷取的資訊；三、文字連結，依中文字結構和直式、橫式兩種書寫方向，設置幾何條件連結候選文字字串；四、候選字串分類，以SVM加入影像中文字的特徵，分類文字字串和非文字字串。使得此方法可以偵測中文和英文兩種語言，並且達到不錯的辨識效果。 / Text messages in an image usually contain useful information related to the scene, such as location, name, direction and warning. As such, robust and efficient scene text detection has gained increasing attention in the area of computer vision recently. However, most existing scene text detection methods are devised to process Latin-based languages. For the few researches that reported the investigation of Chinese text, the detection rate was inferior to the result for English. In this thesis, we propose a multilingual scene text detection algorithm for both Chinese and English. The method comprises of four stages: 1. Preprocessing by bilateral filter to make the text region more stable. 2. Extracting candidate text edge and region using Canny edge detector and Maximally Stable Extremal Region (MSER) respectively. Then combine these two features to achieve more robust results. 3. Linking candidate characters: considering both horizontal and vertical direction, character candidates are clustered into text candidates by using geometrical constraints. 4. Classifying candidate texts using support vector machine (SVM), the text and non-text areas are separated. Experimental results show that the proposed method detects both Chinese and English texts, and achieve satisfactory performance compared to those approaches designed only for English detection.

本篇研究試圖探討半導體產業供應鏈上游的IC設計業者召開法人說明會後，基於我國半導體產業供應鏈緊密連結之特性，同屬半導體供應鏈的其他中、下游製造和封測廠之機構投資人的交易行為是否將受到IC設計業者宣告之法人說明會資訊影響，亦即證明法人說明會資訊在半導體供應鏈中是否具有垂直資訊移轉效果。實證結果發現在法人說明會召開訊息首次見報日時，供應鏈上游公司之法人說明會訊息確實會影響其中、下游公司之機構投資人的持股變化，於宣告好(壞)消息時買進(賣出)，顯示機構投資人藉由其專業團隊和私有資訊能早一般大眾提前調整其交易策略，而此資訊移轉效果也會隨著公司在供應鏈上之距離而逐漸稀釋。此外，結果亦顯示外資由於地緣限制，相較於投信和自營商更會倚賴法人說明會宣告之資訊調整其持股策略，於宣告好(壞)消息時買進(賣出)。 / This research examine the conference call which hold by the IC design companies will transfer useful information to the institutional investors of IC manufacturing and packaging companies in the supply chain downstream. I am interested in if there is a vertical information transfer in the semi-conduct industry. The empirical results show that the conference call information is significantly influence the holding percentage of the institutional investors of the downstream supply chain companies after the information of conference call first reported in the newspapers. The institutional investors will increase the holding percentage after the good news released and vice versa. It is showed that the institutional investors can gather more information before the conference call and adjust their invest strategy in advance. Furthermore, this vertical information transfer effect will dilute by degrees as distance increases. Lastly, the result also shows that the foreign institutional investors more rely on the information released from the conference call to adjust their invest strategy than native institutional investors.

由於網際網路的普及造成資訊量愈來愈大，在資訊的搜尋、整理與閱讀上會耗費許多時間，因此本研究提出一應用文本主題及關係探勘的方法，將多份文件自動生成一篇摘要，以幫助使用者能降低資訊的閱讀時間，並能快速理解文件所欲表達之意涵。 本研究以電影評論文章為例，結合文章結構的概念，將影評摘要分為「電影資訊」、「電影劇情介紹」及「心得結論」三部分，其中「電影資訊」及「心得結論」為透過本研究建置之電影領域相關詞庫比對得出。接著將餘下之段落歸屬於「電影劇情介紹」，並透過LDA主題模型將段落分群，再運用主題關係地圖的概念挑選各群之代表段落並排序，最後將各段落去除連接詞及將代名詞還原為其所指之主詞，以形成一篇列點式影評摘要。 研究結果顯示，本研究所實驗之三部電影，產生之摘要能涵蓋較多的資訊內容，提升了摘要之多樣性，在與最佳範本摘要的相似度比對上，分別提升了10.8228%、14.0123%及25.8142%，可知本研究方法能有效掌握文件之重點內容，生成之摘要更為全面，藉由此方法讓使用者自動彙整電影評論文章，以生成一精簡之摘要，幫助使用者節省其在資訊的搜尋及閱讀的時間，以便能快速了解相關電影之資訊及評論。 / The rapid development of information technology over the past decades has dramatically increased the amount of online information. Because of the time-wasting on absorbing large amounts of information for users, we would like to present a method in this thesis by using text topic and relationship mining for multi-document summarization to help users grasp the theme of multiple documents quickly and easily by reading the accurate summary without reading the whole documents. We use movie reviews as an example of multi-document summarization and apply the concept of article structures to categorize summary into film data, film orientation and conclusion by comparing the thesaurus of movie review field built by this thesis. Then we cluster the paragraphs in the structure of film orientation into different topics by Latent Dirichlet Allocation (LDA). Next, we apply the concept of text relationship map, a network of paragraphs and the node in the network referring to a paragraph and an edge indicating that the corresponding paragraphs are related to each other, to extract the most important paragraph in each topic and order them. Finally, we remove conjunctions and replace pronouns with the name it indicates in each extracted paragraph s and generate a bullet-point summary. From the result, the summary produced by this thesis can cover different topics of contents and improve the diversity of the summary. The similarities compared with the produced summaries and the best-sample summaries raise of 10.8228%, 14.0123% and 25.8142% respectively. The method presented in this thesis grasps the key contents effectively and generates a comprehensive summary. By providing this method, we try to let users aggregate the movie reviews automatically and generate a simplified summary to help them reduce the time in searching and reading articles.

本研究要探討的是在新聞、論壇和社群媒體討論的相關訊息是否真的會影響匯率的運動的假設。對於這樣的研究目標，我們建立了一個實驗，首先以文字探勘技術應用在新聞、論壇與社群媒體來產生與匯率相關的數值表示。接著，機器學習技術應用於學習得到的數值表示和匯率波動之間的關係。最後，我們證明透過檢驗所獲得的關係的有效性的假設。在此研究中，我們提出一種兩階段的神經網路來學習與預測每日美金兌台幣匯率的走勢。不同於其他專注於新聞或者社群媒體的研究，我們將他們進行整合，並將論壇的討論納為輸入資料。不同的資料組合產生出多種觀點，而三個資料來源的不同組合可能會以不同的方式影響預測準確率。透過該方法，初步實驗的結果顯示此方法優於隨機漫步模型。 / This study wants to explore the hypothesis that the relevant information in the news, the posts in forums and discussions on the social media can really affect the daily movement of exchange rates. For such study objective, we set up an experiment, where the text mining technique is first applied to the news, the forum and the social media to generate numerical representations regarding the textual information relevant with the exchange rate. Then the machine learning technique is applied to learn the relationship between the derived numerical representations and the movement of exchange rates. At the end, we justify the hypothesis through examining the effectiveness of the obtained relationship. In this paper, we propose a hybrid neural networks to learn and forecast the daily movements of USD/TWD exchange rates. Different from other studies, which focus on news or social media, we integrate them and add the discussion of forum as input data. Different data combinations yield many views while different combination of three data sources might affect the forecasting accuracy rate in different ways. As a result of this method, the experiment result was better than random walk model.

現今科技日新月異，不斷突破創新，產業環境變動的步調也越來越快，新竄出之金融科技(Finance Technology)的應用，使得許多企業越加注重技術方面的研發創新，尤其，善加運用專利資訊能有效節省研發經費與時間。因此如何有效運用專利是企業維持競爭優勢不可或缺的一環。 有鑑於此，本研究搜集近年各國專利資料庫之專利資料，將資料分為三個時期，並區分申請中與已申請之專利資料，透過文字探勘技術與機會探索分析出金融科技之發展與趨勢，了解各時期詞彙間之關聯性與差異，再搭配視覺化工具KeyGraph，以描繪出金融科技領域之相關詞彙關聯趨勢圖，挖掘未來潛在趨勢。 本研究之結果了解金融科技在各時期的趨勢發展變化與尋求脈絡，以及過去各時期之專利佈局，因而從結果中發現金融科技之發展方向主體為支付領域，許多支付科技接連出現在三個時期中。然而近幾年，其他金融領域如投資、融資、保險、資料分析等也漸漸浮出，從本研究之第三個時期的高頻字詞高達34個可看出，可見金融科技之專利發展佈局已快速從支付領域拓展至其他金融領域。本研究所挖掘出之潛在趨勢顯示了未來金融科技領域中將會有五大重點發展領域，分別為服務整合領域之雲端科技、支付領域之生物辨識與穿戴支付與加密貨幣、資料分析領域之機器學習與人工智慧、信息收集與處理領域之遠程信息處理科技、以及理財投資領域之理財機器人。 期望本研究結果能幫助企業，在面臨新科技不斷衝擊產業，而產業不斷尋求創新發展之下，能夠快速檢閱目前市場趨勢，藉此釐清並改善自身之發展策略，以因應外部環境之變動，提供企業作為金融科技發展之策略參考，也能有助於企業釐清與制定金融科技之投資方向，以擁有持續的競爭優勢。 / Nowadays, with the rapid advancement of information technologies, the changes of business environment and the way to deal with the changes are becoming faster and faster. The development and adoption of new financial technologies has made many enterprises pay more attention to the research and development (R&D) initiatives. Besides, making good use of patent information can effectively save the budget and time of R&D, so how to effectively use patent information is an indispensable part for enterprises to maintain their competitive advantages. This study collected the patent data from the national patent database, and divided the data into three periods, and distinguished the data between the applying and the applied patents. Through the text mining techniques and chance discovery, this study explored the development and trends of financial technology and also aimed to understand the relevance and differences between the major terms in each period. Then, with the visual tool, KeyGraph, this study illustrated the associations between related terms, and proposed the potential future trends based on the graphs. The results of this study help monitor the changes of the trends and financial technology’s development in the three periods, and understand the patent portfolios in each period. This study has found that the main direction of financial technology’s development is the payment field. Many technologies related to payment have successively appeared in the three periods. However, in recent years, other financial areas such as investment, financing, insurance, data analysis and other areas are gradually emerging, since we found 34 high-frequency terms in the third period. This also shows that the development of financial technology’s patent portfolios has expanded from payment to other financial areas. The potential trends of financial technology’s development in this study are five areas, namely, technologies of cloud, biometric and wearable payment and cryptocurrency, machine learning and artificial intelligence, telematics technology, and robo-advisors. It is expected that this study can serve as a reference for the development of financial technology, and help enterprises be able to quickly review their current market trends, clarify and improve their own R&D strategies to respond to the changes in the external environment. Also, it is hoped that the results can help enterprises clarify and develop their own investment directions to maintain competitive advantages.

京都大学 / 0048 / 新制・課程博士 / 博士(教育学) / 甲第19252号 / 教博第179号 / 新制||教||155(附属図書館) / 32254 / 京都大学大学院教育学研究科教育科学専攻 / (主査)教授 田中 耕治, 准教授 西岡 加名恵, 准教授 石井 英真 / 学位規則第4条第1項該当 / Doctor of Philosophy (Education) / Kyoto University / DGAM

京都大学 / 0048 / 新制・課程博士 / 博士(情報学) / 甲第19338号 / 情博第590号 / 新制||情||103(附属図書館) / 32340 / 京都大学大学院情報学研究科社会情報学専攻 / (主査)教授 石田 亨, 教授 田中 克己, 教授 守倉 正博 / 学位規則第4条第1項該当 / Doctor of Informatics / Kyoto University / DFAM

傳統上認為以中文為母語的孩童並不具有「音位覺識能力」(phonemic awareness)，本研究以實驗方式直接測量以中文為母語孩童的「音位覺識能力」，同時探討孩童「音位覺識能力」及「聲母—韻母覺識能力」(onset-rime awareness) 在英文拼字與讀字上所扮演的角色。在本研究中，一百九十二位國小四年級的孩童參與二項「母語音韻覺識測驗」，包括「聲母/韻母異音測驗」(onset/rime oddity test) 和「韻腹/韻尾異音測驗」(nucleus/coda oddity test)。根據上述二項測驗的成績，將孩童分為三組：第一組孩童(共29人)，其「聲母—韻母覺識能力」和「音位覺識能力」皆佳；第二組孩童(共29人)，其「聲母—韻母覺識能力」佳，但「音位覺識能力」差；第三組孩童(共26人)，其「聲母—韻母覺識能力」和「音位覺識能力」皆差。我們接著利用「拼英文假字測驗」和「讀英文假字測驗」來測量三組孩童的英文拼字與讀字的能力，在施測之前，孩童們有八次的機會學會「拼字」與「讀字」兩項測驗所需具備的「字音對應規則」。結果顯示，雖然孩童的「音位覺識能力」在程度上有所不同，但以中文為母語的孩童已具備「音位覺識能力」。此外，在考慮了孩童們「記憶廣度」(digit span)與「英文聽話字彙」 (English receptive vocabulary)的差異後，「聲母—韻母覺識能力」和「音位覺識能力」皆佳的孩童，在拼字的表現上優於「聲母—韻母覺識能力」佳，但「音位覺識能力」差的孩童來，接著「聲母—韻母覺識能力」佳，但「音位覺識能力」差的孩童的拼字表現優於「聲母—韻母覺識能力」和「音位覺識能力」皆差的孩童。最後，「音位覺識能力」佳的的孩童，在讀字的表現上優於「音位覺識能力」差的的孩童，此外，並無證據顯示「聲母—韻母覺識能力」在孩童的讀字能力上扮演著重要的角色。整體而言，本研究的結果支持其他研究的看法，這些研究認為以中文為母語的孩童在母語習得過程中所發展出來的「音韻覺識能力」對於其英文拼字與讀字的能力上有著一定的貢獻，此外，本研究更進一步顯示，這種貢獻會隨著「音韻覺識」程度的不同而有所改變。 / It has been conventionally assumed that Chinese-speaking children do not have phonemic awareness. In this study, Chinese-speaking children’s phonemic awareness was empirically tested and its role, relative to onset-rime awareness, in the acquisition of English spelling and reading abilities was examined. Two L1 phonological awareness tests (i.e., an onset/rime oddity test and a nucleus/coda oddity test) were administered to a total of 192 Chinese-speaking fourth-graders. The children were selected and categorized based on their performances on the two L1 phonological awareness tests: 29 children with good onset-rime awareness and good phonemic awareness, 29 children with good onset-rime awareness but poor phonemic awareness, and 26 children with poor onset-rime awareness and poor phonemic awareness. The three groups of children were then tested on their abilities to spell and read English pseudowords. Before taking the English pseudoword spelling and reading tasks, the children were provided with eight opportunities to master the requisite letter-sound knowledge for the success in spelling and reading the pseudowords. The results showed that Chinese-speaking children demonstrated phonological awareness at the phonemic level, though varying in degree. Considered along with individual differences in digit span and English receptive vocabulary, children with better phonological awareness at both the onset-rime level and the phonemic level performed better in English pseudoword spelling than children with better onset-rime awareness but poorer phonemic awareness, who in turn, performed better than children with poorer phonological awareness at both levels. Finally, children with better phonemic awareness outperformed the other two groups of children with poorer phonemic awareness in pseudoword reading. Onset-rime awareness did not seem to play a significant role in pseudoword reading. These results support and extend other studies suggesting that the acquisition of English spelling and reading abilities in Chinese-speaking children benefits from the phonological awareness obtained during the course of first language acquisition and that the beneficial effect varies with the levels of phonological awareness.

科学研究費補助金 研究種目:基盤研究(C) 課題番号:15520038 研究代表者:神塚 淑子 研究期間:2003-2005年度

市場導向指的是一種辨識市場狀態和顧客需求的企業理論，可以藉由不同的市場情報蒐集來達成。由此，其中的「市場感知能力」和「市場連結能力」能夠幫助企業不斷地修正、改進、創新及再定義原本的市場觀點，並研究市場中所有角色之間的關係。本研究致力於如何以文字探勘、資料探勘及機會發掘等理論創造一個服務的資訊系統來提升這兩個能力。運用本系統可以協助企業在不同市場中找到潛在資源及合作夥伴以共創複合性產品。換句話說，以這種開放關係和不同公司的共創方式更能達到相互曝光、產品與服務互補的效果。而更好的市場導向可以幫助企業間創造出更好的複合式產品並達到利基市場的競爭優勢。 / Market orientation refers to the business philosophy that focuses on identifying and meeting the stated or hidden needs of customers through various approaches of intelligence gathering activities. To this end, two important capabilities, market sensing and market relating, are to enable business to formulate, examine, modify, renovate and redefine their market views and investigate among all players in the market. This study focuses on how to exploit information system as a service to facilitate the acts of the market sensing and market relating capabilities in terms of information technologies of the text mining technique and the chance discovery theory. In addition, this service would help various businesses seek new, necessary and related resources or fellow companies for each other to cooperate to form a complex service product from different markets. In other words, cooperating in the open relationship, different companies attain more possibility to mutually expose their features, product or service, which achieves a marketing for those companies in different markets as well, and better market orientation is also avail business to make better complex product or service for niche market to get competitive advantages.