依時間順序出現之一連串觀測值，通常會呈現某一型態，而根據所產生的 型態可以作為判斷事件發生的基礎。例如，震波形成原因的判斷﹔追查環 境污染源﹔以及在醫學方面，辨識一個正常人心電圖的型態與患有心臟病 的病人其心電圖的型態…等。對於這些問題，傳統之辨識方法常因前提假 設的限制而失去其準確性。在本文中，我們應用神經網路中的逆向傳播演 算法則來訓練網路，並利用此受過訓練的網路來辨別線性時間數列ARIMA 及非線性時間數列 BL(1,0,1,1)。結果發現，網路對於模擬資料中雙線性 係數介於0.2至$0.8$之間的資料有高達$80\%$以上的辨識能力。而在實例 研究中，我們訓練網路來判斷震波形成的原因，其正確率亦高達80\%以上 。同時，我們也將神經網路應用在環境保護方面，訓練網路來判斷二地區 空氣品質的型態。 / A series of observations indexed in time often produces a pattern that may form a basis for discriminatingetween different classes of events. For instance, in theeology, what are the causes of seismic waves? a earthquakesr the nuclear explosions ?in the eathenics, we can use theethod to inquire the source which pollutes the air in somelace, and in the medicine, to distinguish the difference oflectrocardiograms between a health person and an a patient..ect. In this paper, we utilize the back-propagation to trainnetwork and use of the trained networks to judge the linearRIMA(1,0,0) model between the nonlinear BIL(1,0,1,1) model,e can find that the trained network has a good recognitionhose accurate rate is above 80\% for the coefficient of the bilinear model being equal to 0.5 or 0.8. In a living example, we have trained a network to decidehich is the cause of seismic wave, and the trained networkhose accurate rate is larger than 80\%. At the same time, e also applied neural network in environmental protection.

論文提要內容：（共一冊，17770字，分六章） 本研究用雙音節詞來探討不同音段和聲調在臺灣華語的口語詞彙辨識歷程中的重要性。Cohort模型（1978）非常強調詞首訊息的重要性，然而Merge模型（2000）認為訊息輸入和音韻表徵的整體吻合才是最重要的。因此，本研究企圖探索不同音段和詞首詞尾在臺灣華語的口語詞彙辨識歷程中的重要性。然而，聲調的問題並無在先前的模型裡被討論。因此，聲調在臺灣華語的口語詞彙辨識歷程中所扮演的角色也會在本研究中被討論。另外，詞頻效應也會在本研究中被探索。本研究的三個實驗均由同樣的十五名受試者參加。實驗一是測試不同音段在臺灣華語的口語詞彙辨識歷程中的重要性。實驗一操弄十二個雙音節高頻詞和十二個雙音節低頻詞，每一個雙音節詞的每一個音段都分別被噪音擋住。實驗二是在探索詞首和詞尾在臺灣華語的口語詞彙辨識歷程中的重要性。實驗二操弄十二個雙音節高頻詞和十二個雙音節低頻詞。這些雙音節詞的詞首CV或詞尾VG/N都分別被雜音擋住。實驗三操弄二十四個雙音節高頻詞和二十四個雙音節低頻詞。這些雙音節詞的聲調都被拉平到100赫茲。在這三個實驗中，受試者必須聽這些被操弄過的雙音節詞，並且辨認它們。受試者的反應時間和辨詞的準確率都用E-Prime來記錄。實驗結果顯示，傳統的Cohort模型不能被完全支持，因為詞首訊息被噪音擋住的詞仍能被受試者成功的辨識出來。強調聲音訊息和音韻表徵的整體吻合度的Merge模型，比較能解釋實驗的結果。然而，Merge模型必須要加入韻律節點才能處理臺灣華語的聲調辨識的問題。本研究也顯示，雙音節詞的第一個音節的母音在口語詞彙辨識歷程中是最重要的，而雙音節詞的第二個音節的母音是第二重要的。這是因為母音帶了最多訊息，包括聲調。另外，雙音節詞的詞首和詞尾在臺灣華語的口語詞彙辨識歷程中是扮演差不多重要的角色。母音對於聲調的感知是最重要的。詞頻效應也完全表現在臺灣華語的口語詞彙辨識歷程中。 關鍵詞：口語詞彙辨識歷程、臺灣華語、華語聲調、音段、Cohort模型、Merge模型 / The present study investigated the importance of different segments and the importance of tone in spoken word recognition in Taiwan Mandarin by using isolated disyllabic words. Cohort model (1978) emphasized the absolute importance of the initial information. On the contrary, Merge (2000) proposed that the overall match between the input and the phonological representation is the most crucial. Therefore, this study tried to investigate the importance of different segments and the importance of onsets and offsets in the processing of Mandarin spoken words. However, the issues of tone were not included in the previous models. Thus, the importance of tone was also investigated in this study. The issues about frequency effect were also explored here. Three experiments were designed in this study. Fifteen subjects were invited to participate in all three experiments. Experiment 1 was designed to investigate the importance of different segments in Taiwan Mandarin. In experiment 1, 12 high-frequency disyllabic words and 12 low-frequency disyllabic words were selected. Each segment of each disyllabic word was replaced by the hiccup noise. Experiment 2 was designed to investigate the importance of onsets and offsets. In experiment 2, 12 high-frequency disyllabic words and 12 low-frequency disyllabic words were chosen. The CV of the first syllable and the VG/N of the second syllable were replaced by the hiccup noise. Experiment 3 was designed to investigate the importance of Mandarin tones. In experiment 3, 24 high-frequency disyllabic words and 24 low-frequency disyllabic words were selected. The tones of the disyllabic words were leveled to 100 Hz. In the three experiments, subjects listened to the stimuli and recognized them. The reaction time and accuracy were measured by E-Prime. The results indicated that traditional Cohort model cannot be fully supported because words can still be correctly recognized when word initial information is disruptive. Merge model, which proposed that the overall match between the input and the lexical representation is the most important, was more compatible with the results here. However, Merge model needs to include the prosody nodes, so that it can account for the processing of tones in Taiwan Mandarin. In addition, the current study also showed that the first vowel of the disyllabic word is the most crucial and the second vowel of the disyllabic word is the second influential since the vowel carries the most important information, including tones. The results of experiment 2 demonstrated that the onsets and offsets are almost the same important in Mandarin. Furthermore, vowel is the most influential segment for the perception of Mandarin tones. Finally, frequency effect appeared in the processing of Mandarin words. Keywords: spoken word recognition, Taiwan Mandarin, Mandarin tones, segments, Cohort, Merge

本研究提出一項無所不在的學習系統架構改變傳統數位學習之教育傳遞的方式，讓學習者能在任何時間、任何地點透過行動裝置，連結雲端上的學習系統便能自動取得數位化的教學課程及輔助教材等教學資源，並結合RFID標籤之唯一性及無線辨識功能，除了可感測、讀取周遭環境物件資訊以外，更作為學習者個人辨識登入的方法，進而驗證以取得個人所屬的學習課程或學習進度規劃等功能，創造個人化無所不在學習，期許改善目前數位學習在散播教育資源這塊領域，仍有不足之處所造成的數位落差。系統架構本身採用雲端運算作為後端開發技術，將系統、數位學習課程與資料庫分散在各地，以減少單一系統之資源負載量過大及頻寬不足等問題，讓學習者接受即需的教學服務。 / This paper proposes a ubiquitous learning system framework, in order to improve traditional e-learning method for delivering education. Learners could connect to the learning system on the cloud by mobile devices at any time and any place. It could automatically access digital educational resources such as learning courses and supplementary materials. Using the RFID tag for uniqueness and radio frequency identification function. In addition to context-aware and reading the information of objects around the enviroment. It also can be a personalised identify for learners to login and access one’s own function of learning courses or learning scheduling to create personalized ubiquitous learning. There is lack of education due to digital divide which calls for improvement of distributing educational resources. The proposed system architecture incorporates cloud computing for back-end development technology. In order to reduce resource overload of the single system and lack of bandwidth, distributing the learning system, e-learning course and database to several places enable learners to take on-demand educational services.

由於可攜式行動裝置近年來快速成長，隨著行動裝置與地理資訊的結合發展出的適地性服務 (Location-Based Service, LBS)，也呈現越來越多的應用方式。LBS最基本的功能為定位，定位精度好壞也影響著LBS的應用面，越高的定位精度應用面越廣。常見的定位方式便是用GPS導航定位，但是在都會地區越是繁榮的地方往往過多的建物使衛星訊號被阻擋，導致GPS的定位會受到影響，造成平面誤差量會在15公尺以上，或是無法定位。利用都市區域有Wi-Fi訊號分布的特性，藉由區分不同的Wi-Fi訊號來源與接收強度，做出定位判斷，可以使原本GPS無法定位的區域，也可以利用Wi-Fi訊號來做定位。本研究嘗試使用高感度GPS接收器，在建物周邊定位，及在GPS無法運作時使用Wi-Fi訊號來做定位依據，提高整體的定位成功率，並嘗試利用Wi-Fi定位與GPS協同運作來提高整體精度。Wi-Fi定位採用訊號紋辨識法，欲使用訊號紋辨識法必先建立Wi-Fi訊號資料庫，包含地理坐標與對應的訊號來源與強度分布。實驗區為政治大學山下校區綜合院館四周面積約1.76公頃區域，模擬被遮蔽狀況下的GPS導航定位精度，配合Wi-Fi定位來輔助GPS定位結果。結果顯示利用兩種定位系統，可使平面平均定位誤差小於10公尺，高程平均定位誤差小於1.5公尺。 / The positioning accuracy is an important issue for Location-Based Service (LBS). LBS has many products, and its applications are usually based on Global Positioning System (GPS), because GPS navigation has been very mature. GPS has a typical outdoor positioning error of up to 15 meters for civilian users. Hence it has become a viable method for civilian to carry out coarse positioning. However, it has its shortcomings. GPS is available only in outdoors with a clear view of the sky. Since Wi-Fi (Wireless Fidelity) has become another positioning technology which is capable of performing positioning in indoor environments and urban canyons. Wi-Fi positioning is using fingerprinting in this study. This study is desirable to integrate GPS and Wi-Fi positioning technologies for ubiquitous positioning. The related issues of Wi-Fi/GPS technologies, such as database quality, analysis of algorithms, and database processing procedures were studied. Test data sets from National Chengchi University (NCCU) campus will be used to test the proposed algorithms. By using those two positioning system , it is revealed that the positioning accuracy made at the test sites resulted that the 2D coordinate average error is less than 10 meters, elevation positioning average error is less than 1.5 meters.

影像辨識一直是電腦視覺中很重要的技術，且伴隨著行動裝置與相機的普及，人們更加重視辨識的準確度與效能，以區域梯度分佈及直方圖表示方法為基礎的影像特徵描述子，如SIFT與SURF，是近十多年來的物件辨識技術中所採用的主流演算法，然而此類特徵表示法，常需要為多維度的資訊提供大量的儲存空間與複雜的距離計算流程，因此，近年來有學者提出了另一種形式的區域二元特徵描述子 ( Local Binary Descriptor, LBD)，以二元架構建立描述子，使得LBD能在較少空間之下提供可相抗衡的辨識率。 本論文提出以多元編碼機制之區域特徵描述子(LMLED)，乃基於LBD的基本架構，但改以多元編碼取代LBD的二元編碼方法，利用緩衝區的架構達到更強的抗噪性，並提出降維方法以承襲二元編碼在儲存空間的優勢，使得多元編碼機制之區域特徵描述子能在不影響匹配能力與儲存空間的情況下，得到更佳的影像辨識能力。 / Efficient and robust object recognition is an important yet challenging task in computer vision. With the popularity of mobile equipment and digital camera, the demand for effectiveness and efficiency in image recognition has become increasingly pressing. In the past decade, local feature descriptors based on the distribution of local gradients and histogram representation such as SIFT and SURF have achieved a certain level of success. However, these descriptors require a large amount of storage and computing resources for high dimensional feature vectors. Hence, local binary descriptor (LBD) arises and becomes popular in recent years, providing comparable performance with binary structure that needs dramatically lower storage cost. In this thesis, we propose to employ multi-level encoding scheme to replace binary encoding of LBD. The resultant descriptor is named local multi-level encoding descriptor (LMLED). LMLED takes advantage of multiple decision intervals and thus can achieve better noise resistivity. Methods to reduce the dimension have been devised to maintain low storage cost. Extensive experiments have been performed and the results validate that LMLED can achieve superior performance under noisy condition while maintaining comparable matching efficacy and storage requirement.

數位典藏是將物件以數位影像的形式進行典藏，並放置在網路系統供使用者瀏覽，能達到流通推廣與保存維護的效果。但在目前資訊爆炸的時代，數位典藏若僅透過詮釋資料描述是無法有效幫助使用者獲得內容資訊，唯有將之建置成全文檢索模式，才能方便使用者快速檢索到所需資訊，而光學字元辨識技術（簡稱OCR）能協助進行全文內容的輸出。 本研究藉由實際操作OCR軟體辨識明代古籍，探究古籍版式及影像對於軟體辨識結果之影響；藉由深度訪談訪問有實際參與數位典藏全文化經驗之機構人員，探究機構或個人對於計畫施行之觀點與考量。結果發現，雖然實際辨識結果顯示古籍版式與影像會對於OCR辨識有所影響，綜合訪談內容得知目前技術層面已克服古籍版式的侷限，但對於影像品質的要求仍然很高，意指古籍影像之品質對OCR的辨識影響程度最大；雖然OCR辨識技術已經有所突破，顯示能善用此技術協助進行全文資料庫的建立，但礙於技術陌生、經費預算、人力資源等因素，使得多數機構尚未運用此技術協助執行數位典藏全文化。 本研究建議，機構日後若有興趣執行數位典藏全文化計畫，首先，需要制定經常出適合機構執行的作業流程，並且瞭解自身欲處理物件之狀況，好挑選出適合的輸入處理模式；再者，需要多與技術廠商溝通協調，瞭解所挑選之物件是否符合處理上的成本效益；最後，綜合典藏機構與使用者之需求考量下，建議未來採取與OCR廠商合作的方式，由使用者自行挑選需要物件進行OCR辨識，校對完成後將全文內容回饋給典藏機構。這樣不僅能瞭解使用者需求為何，也能降低機構全文校對所耗費的成本。 / Digital Archives, placed in the network system for users to browse, change the collection into the digital images, and can help to preserve the collection and promote the content information. However, in the era of information explosion, Digital Archives can’t help users to retrieve the information in the collection by simply recording metadata. So, only when built into the full text retrieval can Digital Archives provide users with a quick retrieval of the information they want. And the Optical Character Recognition (OCR) can help to output the full text information. The study explores the ancient books’ format and impact of image quality on the recognition results by recognizing the ancient books of the Ming dynasty with the OCR software. The study also explores institutional as well as individual views and considerations by in-depth interviewing institutional staff with experiences in the full text of Digital Archives plan. From the result we can discover that though the ancient books’ format and image quality do have influences on the recognition results, the overall interview suggests that the technology has overcome the limitation of the format under the high requirement for the image quality; that is, the quality of ancient books’ images is the most influential factor in the recognition results. Although the OCR already has the breakthrough in assisting the establishment of the full text database, most institutions have not yet applied this technology to full-textualization of the Digital Archives due to technical unfamiliar, budget, human resources and other factors. The study suggests that if some day one institution is interested in working on the the full text of the Digital Archives project, it firstly needs to develop a proper SOP and needs to understand the conditions of their ready-to-be-textualized collections so that it can adopt a suitable input mode. Secondly, this institution needs to communicate with the OCR company more so that it can realize whether the chosen collection fits the cost-effectiveness. Finally, under the considerations of both the institution and users, the study suggests that institutions can cooperate with OCR companies in the future, so users can choose collections for OCR recognition on their own and give the full text to the institutions as feedback after proofreading. This can not only understand users’ needs but also reduce the cost of the proofreading for the institution.

每個人遇到曾經看過的物體時，辨識該物體速度會增加（或辨識的正確率增加），這個現象稱為促發(priming)效果（簡稱P-P促發），同樣地，閱讀某物體的名稱（亦即文字）後，隔幾分鐘後再看該物體的圖形，這樣也會產生一種促發量（簡稱W-P促發）。許多研究都指出W-P促發是一種內隱(implicit)記憶，亦即，個體不需要刻意想起曾經看過的物件，促發效果仍會產生，而且P-P促發量都高於W-P促發量。然而，一些研究卻發現W-P促發量等於P-P促發量，顯然地，內隱記憶理論無法對於這種反直覺現象提出合理的解釋。 根據Paivio的雙重收錄理論(dual coding theory)(Paivio, 1986, 1991)，辨識具體(concrete)名詞（例如，物體的名稱）會同時觸及(access)或激發兩種知識表徵，一種是涉及左腦的口語(verbal)表徵，另一種是涉及左腦與右腦的影像(image)表徵，而許多神經語言學研究皆指出，涉及處理具體名詞的神經機制不只包含左腦，同時也包含右腦，是以，閱讀具體名詞可能會觸及或激發物體的內在表徵，如果物體輪廓相當獨特或明顯，那麼閱讀此類型物體之名稱可能會觸及或激發此類物體的完整或重要表徵，致使W-P促發量等於P-P促發量現象。因此本研究試圖操弄物體輪廓診斷性來解釋W-P促發量等於P-P促發量之現象。 實驗一與實驗二分別以「圖形唸名」以及「圖形知覺辨識作業」來檢驗「物體輪廓診斷性」對促發的影響，結果顯示，「整體診斷性不高」物體（globally non-diagnostic object，簡稱GN類物體）的P-P促發量高於W-P促發量，和先前許多研究結果一致，然而，「整體診斷性高」物體（globally diagnostic object，簡稱GD類物體）的W-P促發量等於P-P促發量，顯示「物體輪廓診斷性」會影響促發的表現，同時也顯示閱讀GD類物體名稱可以觸及或激發GD類物體的整體或必要的知識表徵。 實驗三以分視野(divided visual field)呈現方式檢驗GD類物體的W-P促發之腦側化現象。本研究發現，顯著的W-P促發只出現在右腦，顯示W-P促發主要經由右腦來處理，根據Paivio的雙重收錄理論來推論，W-P促發之本質可能主要涉及以影像為基礎的(image-based)的知識表徵。 本研究同時操弄外顯記憶以檢驗外顯記憶是否污染W-P促發而導致W-P促發量等於P-P促發量，結果顯示，不管哪一種物體，P-P情境的再認記憶表現都顯著比W-P情境好，顯示GD類物體的促發表現與外顯記憶表現之間有單一分離(single dissociation)的關係，換言之，GD類物體之W-P促發並不受外顯記憶影響或污染。此外，實驗四顯示刻意的心像策略並不涉及W-P促發，顯示閱讀GD類物體名稱觸及GD類物體概念表徵是一種自動化而且相當快速的歷程。 / Implicit memory is usually assessed by showing repetition priming effects, when better performance in accuracy or response time for stimuli that have been previously encountered in comparison with performance with new stimuli. Picture-naming priming has been examined in studies that compared priming in participants who named pictures in the study phase and named those same pictures in the test phase (P-P condition) versus participants who read words that were the names of pictures in the study phase and named pictures cor-responding to those words in the test phase (W-P condition). Many studies demonstrated W-P priming is less than P-P priming in the picture-naming task and other similar object recognition tasks. However, in sharp contrast to the above studies, some studies reported equivalent magnitudes of P-P and W-P naming priming. Theories of implicit memory cannot account for the counter-intuitive phenomenon. According to Paivio’s dual-coding theory, the processing of abstract nouns (e.g., justice) relies on verbal code representations of the left cerebral hemisphere only, whereas concrete nouns (e.g., airplane) additionally access a second image-based processing system in the right cerebral hemisphere (Paivio, 1986, 1991). Paivio’s theory is supported by many researches on neurolinguistics. If the contour of an object is very distinctive or diagnostic, there should be the possible result that reading the name of the distinctive objects could access the whole or essential representation of the object. Following the idea, I manipulated global diagnosticity of object contour to examine whether P-P priming is always larger than W-P priming. I found P-P priming was equivalent to W-P priming on “globally diagnostic” (GD) objects, but the P-P priming was still larger than W-P priming on “glob-ally non-diagnostic” (GN) objects. This phenomenon appeared on both pic-ture-naming (Experiment 1) and picture perceptual-identification (Experiment 2) tasks. Experiment 3 showed that significant W-P priming appeared only when GD objects in the test phase were presented to the right cerebral hemi-sphere (in the left visual field). Based on the Paivio’s dual coding theory (Paivio, 1986, 1991) and research on neurolinguistics, the nature of W-P priming for GD objects was inferred to be image-based processing. Better explicit (conscious) memory performance (recognition memory) in P-P condition than that in W-P condition showed that equivalent priming across P-P and W-P conditions on GD objects was dissociated from the influence of conscious recognition memory. Experiment 4 showed that the intentional strategy of generating mental imagery was not necessarily involved in the W-P priming. These results suggested that reading names of globally diagnostic objects can access, automatically and unconsciously, the representation or essential features of globally diagnostic objects, and right cerebral hemisphere might be responsible for the processing.

在音樂中，組曲是一種特別的創作形式。組曲將多首音樂段落組合排列，並且在音樂段落之間加入間奏，形成一首音樂組曲。組曲的編曲重點在於音樂段落的編排順序及段落之間的連結。平時在宴會、舞會、餐廳、賣場等場合中，往往都會連續播放多首流行音樂。利用電腦編曲自動產生流行音樂組曲，將可提升播放音樂的銜接與流暢感。 因此，本研究利用資料探勘技術及音樂編曲理論，將多首音樂重新改編成一首組曲。系統首先將每首音樂分段並找出每首音樂的代表段落。接著，系統根據代表段落間的相似度編排順序。最後，為了達到組曲中音樂段落連接的流暢性，我們以模型訓練的方式在段落連結間加入間奏。系統從訓練資料學習產生旋律發展、和弦進程與節奏的模型，接著分析代表段落的動機、旋律、和弦及節奏，使得組曲編曲後的段落連結更為流暢且完整。本研究以流行音樂鋼琴伴奏曲為測試資料，我們分別邀請三十四位受過音樂訓練與未受音樂訓練的測試者，針對本研究所提出的鋼琴伴奏節奏辨識、代表段落萃取、段落順序編排及間奏產生，評估其效果。實驗結果顯示，本研究所提出的順序編排與間奏產生技術，對於組曲的流暢感，有著相當大的幫助。 / In music, a medley is an organized piece composed from segments of existing pieces. Ordering and bridge for connection between segments are the important elements for medley arrangement. Automatic medley arrangement is helpful for playing a set of music continuously which is common in banquet, party, restaurant, shopping mall, etc.. This thesis aims to develop the automatic medley arrangement method by using data mining techniques and music arrangement theory. The proposed method first segments each music and discovers the significant segment from each music. Then, the linear arrangement based on the similarities between significant segments is generated. Finally, in order to connect segments smoothly in the medley, the bridge between two segments is generated and inserted by using model training. Three models, melody progression, chord progression and rhythm models are learned from training data. For the experiments, testing data is collected from popular piano music and thirty-four people are invited to evaluate the effectiveness of the rhythm recognition of accompaniment, the extraction of significant segment, the linear arrangement of segments, and the creation of bridge. Experimental results show that the proposed medley arrangement method performs well.

本研究旨在調查電腦輔助英語發音學習軟體 「MyET」，對學習者在學習英語發音方面的影響。 利用電腦輔助英語發音學習軟體(CAPT)，練習英語的類化效果，也列為調查重點之一。 此外，學生使用CAPT過程中遭遇的困難和挑戰，以及互動過程中發展出來的對策也一一加以探討。 本研究的目的是要把CAPT在英語聲韻教學的領域中做正確的定位，並且探討如何使用其他的中介工具(例如人類)來強化此類軟體的輔助學習效果。 參與本次研究的大學生一共有九十名，分為三組：兩組CAPT組(亦即實驗組，使用CAPT獨自或與同儕一起使用CAPT學習英語發音)、非CAPT組(控制組)一 組。每組三十名。實驗開始，所有學生以十週的時間練習朗讀 從「灰姑娘」(Cinderella) 摘錄的文字，此段文字由發行 MyET 的公司線上免費提供。 實驗前與實驗後，兩組的學生各接受一次測驗。 每週練習結束後，學生必須將學習心得記載於學習日誌上；教師也針對每個學生的學習心得給予指導回饋。 研究結果顯示，兩個CAPT組別(亦即使用CAPT發音學習軟體的組別)的學生在學習英語聲韻的過程中，都有明顯及正面的進步與改變。尤其是語調與速度快慢方面的進步遠勝於發音的進步。再者，實驗組學生以十週的時間利用CAPT學習英語後，在朗讀新的文字時，無論是發音或語調都有類化的效應，但是在速度快慢方面則無顯著進步。然而，實驗結果三組的發音表現，在量化統計上並未達到明顯的差異。 雖然如此，在質化的探究上，經過分析學生的學習心得後得知：所有組別當中，獨自使用CAPT學習英語發音的組別，最能夠自我審視語言學習歷程 (包括模仿和學習樂趣)。至於共同使用CAPT學習的學生自述在英語流暢度、語調及發音方面獲致最大的改善。控制組的學生因為沒有同儕的鷹架教學及回饋，也沒有 MyET提供的練習回饋，練習過程中，學生自述學習困難的頻率最高，學生也認為學習收穫很少。 參與本次研究實驗組的學生認為， CAPT提供練習回饋的機制設計有改進的空間。 有關本研究結果在理論及英語教學上的意涵以及研究限制，於結論當中一一提出加以討論。 關鍵字：電腦輔助語言教學，語音辨識軟體，超音段，語調，時長，學習策略， 中介 / This present study investigated the impact of computer-assisted pronunciation training (CAPT) software, i.e., MyET, on students’ learning of English pronunciation. The investigation foci included the generalization of the effect of practice with the CAPT system. Also examined are the difficulties and challenges reported by the students who employed the CAPT system and the strategy scheme they developed from their interaction with the system. This study aimed to position the role of the CAPT system in the arena of instruction on English pronunciation and to investigate how other kinds of mediation, such as that of peer support, could reinforce its efficacy. This study involved 90 Taiwanese college students, divided into two experimental groups and one control group. The two experimental groups practiced English pronunciation by using a computer-assisted pronunciation training (CAPT) program either independently or with peers while the control group only had access to MP3 files in their practice. All the groups practiced for ten weeks texts adopted from a play, Cinderella, provided by MyET free of charge on line. They all received a pretest and a posttest on the texts they had practiced and a novel text. Each week after their practice with the texts, the participants were asked to write down in their learning logs their reflections on the learning process in Chinese. In the same way, the instructor would provide her feedback on the students’ reflections in the logs every week. The results showed that the ten-week practice with the CAPT system resulted in significant and positive changes in the learning of English pronunciation of CAPT groups (i.e., the Self-Access CAPT Group and the Collaborative CAPT Group). The progress of the participants in intonation and timing was always higher than in segmental pronunciation. Moreover, the ten-week practice with the CAPT system was found to be generalized (though the generalization is less than mediocre) to the participants’ performance in the production of segmental pronunciation and intonation but not in the timing component in reading the novel text. However, the improvement of the CAPT groups was not great enough to differentiate themselves from the MP3 Group. Though the quantitative investigation did not reveal significant group differences, the qualitative analysis of the students’ reflections showed that the learning processes all the three groups went through differed. The Self-Access CAPT Group outperformed the other two groups in developing self-monitoring of language learning and production, and in enjoying working with the CAPT system/texts. Among the three groups, the Collaborative CAPT Group outscored the other two groups in reporting their gains and improvement in fluency, intonation and segmental pronunciation, as well as developing strategies to deal with their learning difficulty. Though the students in the MP3 group also made significant progress after the practice, without peers’ scaffolding and the feedback provided by MyET, they reported the highest frequency of difficulties and the least frequency of gains and strategies during the practice. The participants of this study also considered necessary the improvement of the CAPT system’s feedback design. At the end of the study theoretical and pedagogical implications as well as research limitations are presented. Key words: Computer-Assisted Language Learning (CALL), Automatic Speech Recognition System (ASRS), segmental pronunciation, prosody, intonation, timing, learning strategies, mediation

本論文以語意學、認知語言學及心理語言學的觀點，探討中文名詞的多義現象。本篇論文討論詞彙多重意義的語言及心理語言表徵，並研究詞義如何在心理詞彙庫中被擷取。 文中藉由認知語言學者Tuggy（1993）所提出的認知語意架構，及詞彙語意學中「同形異義詞（homonymy）」、「一詞多義（polysemy）」及「模糊意義（vagueness）」等觀念之區辨，探索詞義之本質。本文所談之意義（sense）根據Ahrens（1999）之定義有以下三個特性：（1）意義非轉喻（metonymy）或部分/全體（meronymy）之延伸，但可以是隱喻性延伸；（2）意義之間的延伸關係無法由在同類名詞中直接以規律獲得；（3）除非特意，一詞的不同意義不會在同一語境中同時出現。 針對詞彙辨識歷程的語意效應，本論文有三大研究主題--意義數目效應、意義相關性效應及相對意義頻率效應。各效應之預測如下：「意義數目效應」認為意義數目較多的詞彙在詞彙判斷作業（lexical decision tasks）中比較容易辨識；「意義相關性效應」認為當一個詞的意義之間的相關性較高時，該詞的辨識時間亦會較快；「相對意義頻率效應」則認為一個詞意義之間頻率差異程度越小，該詞越有歧義之特質，也因此越容易辨識。 本研究旨在以實證方式探討這些語意向度在詞彙辨識歷程中所扮演的角色。由受試者提供200個中文名詞的意義並決定意義間的相關性；進行電腦詞彙判斷作業，得到辨識每個詞所需的反應時間；再由統計考驗反應時間來驗證各效應。 本研究結果發現只有「意義數目效應」達顯著水準，意義數目較多的詞彙在詞彙判斷作業中比較容易辨識。此一效應支持心理詞彙庫的隨機觸接模型（random access model）及平行觸接模型（parallel access model）並駁斥「序列搜尋模型（serial access model）」。 / This thesis studies the multiple senses of Chinese nominals from semantic, cognitive linguistic, and psycholinguistic viewpoints. It discusses the linguistic and psycholinguistic representations of a word's multiple senses, and the access of these senses in the mental lexicon. The nature of lexical meaning is examined by discussing lexical semantic notions such as homonymy, polysemy, and vagueness, and by introducing Tuggy's (1993) cognitive linguistic representation of a word's multiple lexical meanings. The "senses" of a word are defined according to the lexical semantic theory of Ahrens (1999) as having three properties: (1) A sense is not an instance of metonymic or meronymic extension, but may be an instance of metaphorical extension. (2) The extension links between two senses cannot be inherited by a class of nouns. (3) Senses cannot appear in the same context (unless the complexity is triggered). This thesis specifically looks into three semantic effects on word recognition, including the word's number-of-sense (NOS) effect, the effect of sense relatedness, and the effect of relative sense frequency. The predictions of these effects are as follows: The NOS effect predicts that words with more senses are recognized faster than those with fewer senses. The effect of sense relatedness predicts that words with more closely related senses are easier to recognize than those with distantly related senses. The effect of relative sense frequency predicts that words with equal sense frequencies are more easily recognized than words with unequal sense frequencies. This research aims at empirically verifying these three effects during the recognition of isolated lexical items. Subjects generate the senses of 200 Chinese nominals, and rate the relatedness among these senses. Lexical decision tasks are conducted to obtain the reaction times required to recognize each stimulus item. The semantic effects are verified by comparing the reaction times of different groups of experimental stimuli. The experimental results confirm only the NOS effect, giving support to the random and parallel access models of lexical access, and refuting the serial access model of the mental lexicon.