No description available.

abstract: High-level inference tasks in video applications such as recognition, video retrieval, and zero-shot classification have become an active research area in recent years. One fundamental requirement for such applications is to extract high-quality features that maintain high-level information in the videos. Many video feature extraction algorithms have been purposed, such as STIP, HOG3D, and Dense Trajectories. These algorithms are often referred to as “handcrafted” features as they were deliberately designed based on some reasonable considerations. However, these algorithms may fail when dealing with high-level tasks or complex scene videos. Due to the success of using deep convolution neural networks (CNNs) to extract global representations for static images, researchers have been using similar techniques to tackle video contents. Typical techniques first extract spatial features by processing raw images using deep convolution architectures designed for static image classifications. Then simple average, concatenation or classifier-based fusion/pooling methods are applied to the extracted features. I argue that features extracted in such ways do not acquire enough representative information since videos, unlike images, should be characterized as a temporal sequence of semantically coherent visual contents and thus need to be represented in a manner considering both semantic and spatio-temporal information. In this thesis, I propose a novel architecture to learn semantic spatio-temporal embedding for videos to support high-level video analysis. The proposed method encodes video spatial and temporal information separately by employing a deep architecture consisting of two channels of convolutional neural networks (capturing appearance and local motion) followed by their corresponding Fully Connected Gated Recurrent Unit (FC-GRU) encoders for capturing longer-term temporal structure of the CNN features. The resultant spatio-temporal representation (a vector) is used to learn a mapping via a Fully Connected Multilayer Perceptron (FC-MLP) to the word2vec semantic embedding space, leading to a semantic interpretation of the video vector that supports high-level analysis. I evaluate the usefulness and effectiveness of this new video representation by conducting experiments on action recognition, zero-shot video classification, and semantic video retrieval (word-to-video) retrieval, using the UCF101 action recognition dataset. / Dissertation/Thesis / Masters Thesis Computer Science 2016

The present study aims to explore the relationship between proficiency level and semantic retrieval in the second language. A group of Korean bilinguals who speak English with high proficiency performed semantic relatedness judgement tasks of two hundred English word pairs. Unbeknownst to the participants, half of the words in both the related and the unrelated categories contained a "hidden prime"—a common first syllable shared by the two words, if translated into Korean. Each participant's event-related potential (ERP) was recorded while reading the words. While a former study by Thierry and Wu (2007) found that Chinese-English bilinguals were affected by the hidden primes, thus causing a "N400 reduction effect" in their averaged ERP, the bilingual group of the present study was unaffected by the hidden primes. The difference between the bilingual groups' performance between Thierry and Wu's study and the present study is likely caused by the higher English proficiency of the bilingual group in the present study. This provides additional evidence supporting the Revised Hierarchical Model of semantic retrieval proposed by Kroll and Steward (1994), which suggests that increased proficiency leads to reduced reliance on the first language during second language semantic retrieval.

Afin d’assurer la fabrication de produits conformes et à faible coût dans les industries, la maîtrise des procédés de fabrication est devenue un enjeu majeur. Les systèmes d'information dans les industries sont assez complexes et les besoins métier évoluent en permanence, rendant ainsi difficile la recherche des ressources qui fournissent les informations manufacturières pour le contrôle de procédés industriels. De plus, l’utilisation de plateformes logicielles commerciales dans les industries pour le traitement des données ne facilite pas l’accès à l’information produite car ces plateformes ne permettent pas la gestion sémantique de l’information. Cette thèse défend l’idée qu’il faut réduire la distance entre les ressources disponibles et les besoins métier des experts qui assurent le contrôle de processus industriels. L’approche S3 est proposée pour permettre à la fois la description et la recherche de ces ressources. S3 repose sur deux stratégies de recherche complémentaires: une stratégie ascendante permettant la création de descripteurs sémantiques de ressources, et une stratégie descendante permettant la capture des besoins métier dans des patterns de recherche. Deux structures sémantiques sont proposées pour supporter les mécanismes de description et de recherche: une ontologie « manufacturing process » et un dictionnaire « process control ». Chaque stratégie de recherche, appuyée par les structures sémantiques apporte un niveau de description différent et permet l’alignement de différents types de connaissances métier. Cette approche a été expérimentée au sein de l’entreprise STMicroelectronics et a révélé des résultats prometteurs. / In order to ensure the manufacturing of conforming products with the least waste, the manufacturing process control has ever more become a major issue in industries nowadays. The complexity of the information systems in industries and the permanent evolution of the business needs make difficult the retrieval of the resources that provide manufacturing information related to the process control. In addition, the use of commercial software platforms in industries for the processing of data, does not facilitate the access to the information produced, because these platforms do not support the semantic management of information.This thesis argues the need to reduce the distance between the used resources in industries and the business needs of the experts that ensure the control of the manufacturing processes.The S3 approach is proposed to support the control of the manufacturing processes through an original resource management system. This system is intended for both resource description and retrieval. The S3 approach relies on two complementary retrieval strategies: a bottom-up strategy enabling the creation of semantic descriptors of resources, and a top-down strategy enabling the capture of business needs in search patterns. Two semantic structures are proposed to support the resource description and retrieval mechanisms: a manufacturing process ontology and a process control dictionary. Basing on these semantic structures, each retrieval strategy provides different levels of description to the resources, and enables the alignment of different types of business knowledge. The experimentation of the approach within STMicroelectronics showed promising results.

Adaptive reuse of parametric finite element analysis (FEA) models is a common form of reuse that involves integrating new information into an archived FEA model to apply it towards a new similar physical problem. Adaptive reuse of archived FEA models is often motivated by the need to assess the impact of minor improvements to component-based designs such as addition of new structural components, or the need to assess new failure modes that arise when a device is redesigned for new operating environments or loading conditions. Successful adaptive reuse of FEA models involves reference to supporting documents that capture the formulation of the model to determine what new information can be integrated and how. However, FEA models and supporting documents are not stored in formats that are semantically rich enough to support automated inference of their relevance to a modelers needs. The modelers inability to precisely describe information needs and execute queries based on such requirements results in inefficient queries and time spent manually assessing irrelevant models. The central research question in this research is thus how do we incorporate a modelers intent into automated retrieval of FEA models for adaptive reuse? An automated retrieval method to support adaptive reuse of parametric FEA models has been developed in the research documented in this thesis. The method consists of a classification-based retrieval method based on ALE subsumption hierarchies that classify models using semantically rich description logic representations of physical problem structure and a reusability-based ranking method. Conceptual data models have been developed for the representations that support both retrieval and ranking of archived FEA models. The method is validated using representations of FEA models of several classes of electronic chip packages. Experimental results indicate that the properties of the representation methods support effective automation of retrieval functions for FEA models of component-based designs.

Thanks to the advanced technologies and social networks that allow the data to be widely shared among the Internet, there is an explosion of pervasive multimedia data, generating high demands of multimedia services and applications in various areas for people to easily access and manage multimedia data. Towards such demands, multimedia big data analysis has become an emerging hot topic in both industry and academia, which ranges from basic infrastructure, management, search, and mining to security, privacy, and applications. Within the scope of this dissertation, a multimedia big data analysis framework is proposed for semantic information management and retrieval with a focus on rare event detection in videos. The proposed framework is able to explore hidden semantic feature groups in multimedia data and incorporate temporal semantics, especially for video event detection. First, a hierarchical semantic data representation is presented to alleviate the semantic gap issue, and the Hidden Coherent Feature Group (HCFG) analysis method is proposed to capture the correlation between features and separate the original feature set into semantic groups, seamlessly integrating multimedia data in multiple modalities. Next, an Importance Factor based Temporal Multiple Correspondence Analysis (i.e., IF-TMCA) approach is presented for effective event detection. Specifically, the HCFG algorithm is integrated with the Hierarchical Information Gain Analysis (HIGA) method to generate the Importance Factor (IF) for producing the initial detection results. Then, the TMCA algorithm is proposed to efficiently incorporate temporal semantics for re-ranking and improving the final performance. At last, a sampling-based ensemble learning mechanism is applied to further accommodate the imbalanced datasets. In addition to the multimedia semantic representation and class imbalance problems, lack of organization is another critical issue for multimedia big data analysis. In this framework, an affinity propagation-based summarization method is also proposed to transform the unorganized data into a better structure with clean and well-organized information. The whole framework has been thoroughly evaluated across multiple domains, such as soccer goal event detection and disaster information management.

教育部自九十年度起實施九年一貫英語教學課程，由於學生程度不一，班級人數過多，教學媒體不足，研究報告指出應該成立英語教學資源中心，提高教師教學品質，規劃線上輔助教學，以增進學生學習機會及學習效果。 本研究從知識管理角度探索英語線上輔助教學，建構英語教學內容知識結構模式及教學知識流程架構；俾使線上教學具備語意查詢機制，開啟英語線上輔助教學創新模式。 英語教學內容知識框架（English Content Knowledge Ontology）抽象化類別具多重之階層、對等及相依關係。將內容知識抽象化類別實體化，其類別與實體之集合，即為本範例所建構之知識框架。 本研究引介物件導向，視內容知識為一個物件，具有不同之狀態及行為；將狀態與行為封裝，物件與物件間藉由訊息進行協調動作，達成整體運作之功能。學生亦為學習流程中一個物件，具有不同之屬性及行為；透過訊息之交換、進行任務協調，達成學習之最終目的。 以UML表示法建構英語線上學習流程三大模型：功能模型─使用案例模型將使用者對系統的需求模型化，靜態模型─物件模型抽出物件，表現出靜態的結構，動態模型─表示出物件與物件之間訊息的流向。 研究中試圖探討有關國小英語線上輔助教學網站之現況。未來期望架構具備語意檢索機制之網站，採自由選擇學習、多元適切評量模式，能紀錄分析學習及測試結果，建立學習紀錄（包括學習風格、習性、態度及學習效果等等）以期增進學生學習機會，分享教學資源。 藉由新的學習模式，學生可以獲得自我成長的資源，同時希望可以喚起更多關懷英語教學網站之研究與發展，為我們的孩子及新的學習方式注入更多生命力。本研究所建構之模式，希望能作為未來發展兒童英語線上學習之參考，以提昇國內兒童英語教育實施之成效。 關鍵詞：九年一貫英語課程、知識管理、知識本體、語意資訊檢索、物件導向、UML / The Grade 1-9 English Teaching Curriculum was put into practice by the Ministry of Education in 2001. Due to the fact that there are too many students in each class and that there is not enough teaching media, the students fall into different levels in English proficiency. Some research concluded that an English Teaching Resource Centre should be established to enhance teachers’ teaching quality and to provide an on-line teaching aid, hence to give the students more learning opportunities and to improve their learning effects. In this study, the on-line English teaching aid was discussed from the viewpoint of Knowledge Management. English Content Knowledge Ontology and English Teaching Knowledge Flow Architecture were suggested to provide an on-line teaching aid with a semantic retrieval system so as to initiate an innovative model for on-line English teaching aids. The English Content Knowledge, Ontology, a model of English domain knowledge, defines the concepts and their attributes, as well as the multiple relationships between the concepts: Class（vertical）, Reciprocity（horizontal）, and dependency（grouping）relationships. An “instance” is hypostatized from an abstract class, and the integration of the abstract classes and the instances represent the English Content Knowledge Ontology built up in this research. The present study has introduced Object Oriented concept that deemed the English content knowledge as an object and as an instance with different Attributes and Operations. Encapsulated attributes and operations engaged in coordination among different objects through exchanges of messages, and resulted in the achievement of the overall system operations. Meanwhile, each student would also be an object in the course of the study and possess different attributes and operations. The final goal, learning, could be achieved through the exchanges of messages for mission coordination. UML was applied in the study to construct the three major models in the on-line English learning. Functional Model – Use case Diagram to model the user’s requirements for the system； Static Model – Class Diagram to abstract an object for showing a static architecture； Dynamic Model – Sequence Diagram to describe the information flow among the objects. This research also attempted to explore the present circumstances of the primary school’s on-line English teaching aid websites. It illustrated the needs of developing architecture for a website with semantic retrieval functions, multiple choice ways of learning and diversified modes for learning assessments It will be able to record and analyse the learning effect and test the result, building up learning records (including learning style, habit, attitude and learning effect) in order to provide each student with more learning opportunities and to share teaching resources with all of the teachers. Through the new learning model, the students are able to obtain learning resources to grow by themselves. The present study also urged more concern on the research and development of the English teaching website in order to provide more vitality to our children and their new learning methods. The model built up in this study may serve as a reference in the development of effective on-line English learning for children. Key words: Grade 1-9 English Teaching Curriculum, Knowledge Management, Ontology, Semantic Retrieval System, Object Oriented, UML