Conventional pattern recognition systems have two components: feature analysis and pattern classification. For any object in an image, features could be considered as the major characteristic of the object either for object recognition or object tracking purpose. Features extracted from a training image, can be used to identify the object when attempting to locate the object in a test image containing many other objects. To perform reliable scene analysis, it is important that the features extracted from the training image are detectable even under changes in image scale, noise and illumination. Scale invariant feature has wide applications such as image classification, object recognition and object tracking in the image processing area. In this thesis, color feature and SIFT (scale invariant feature transform) are considered to be scale invariant feature. The classification, recognition and tracking result were evaluated with novel evaluation criterion and compared with some existing methods. I also studied different types of scale invariant feature for the purpose of solving scene analysis problems. I propose probabilistic models as the foundation of analysis scene scenario of images. In order to differential the content of image, I develop novel algorithms for the adaptive combination for multiple features extracted from images. I demonstrate the performance of the developed algorithm on several scene analysis tasks, including object tracking, video stabilization, medical video segmentation and scene classification.

A remarkable thing about natural language is that we can use it to share our beliefs and thoughts about the world with other speakers of our language. In cases of successful communication, beliefs seem to be transferred from speakers to hearers by means of the hearer recovering the contents of the speaker’s utterances. This is so natural to us that we take it for granted in our everyday life, and rarely stop to think about how it's is possible. Nevertheless, it's a phenomenon that calls for explanation. It is natural to expect that natural language semantics has a key explanatory role to play here. In order to understand this role, we must relate the semantic values assigned to sentences by semantic theories with the contents of our speech acts. The simplest possible relation would be identity; the meaning of a sentence is simply the belief expressed by an assertion of the sentence in a given context of utterance. However, a number of problem cases in the literature suggest that this cannot be the case. This dissertation offers a critical assessment of the arguments for distinguishing the semantic value of a sentence from its so-called assertoric content, focusing on problems arising from the analysis of tense and temporal expressions. I conclude that they are indeed distinct, and offer a constructive account of how they must be related in order to allow for an explanation of communicative success.​

This research contributes to the field of ontology-based semantic matching techniques and also to the field of Instant Messaging (IM) based enhanced presence. It aims to achieve a mutually beneficial development of two fields through interactions in their use of data and their functionality. With respect to semantic matching this research has developed a collaborative and self-evolutionary approach based on user involvement in order to overcome disadvantages of traditional ontology-based approaches. At the same time, enhanced semantic matching algorithms were also explored and developed to achieve better performance when searching and querying through the ontology. In order to realize this automatic, dynamic and collaborative approach, a Jabber-based IM system was built to support its development with specific data and to evaluate its performance. In the prototype of the system, Computer Science area is selected to be the domain of the ontology in order to demonstrate the practicability of the new approach. With respect to enhanced presence an efficient semantic-based contacts search engine which can feature context-based search ranking is provided to support academic researchers. It is especially designed to help new academic researchers to find potential contacts who share a common research interest. It enriches the IM system's presence information, and helps the user to pick the most suitable contacts and conveniently organize meetings or co-operating with others. Consequently, this research improves the efficiency of users' academic researching, and extends users' relationship radius during their academic research careers. The contributions are particularly highlighted by the comprehensive support during the academic user's self-educational process.

Natural Language Processing systems crucially depend on the availability of lexical and conceptual knowledge representations. They need to be able to disambiguate word senses and detect synonyms. In order to draw inferences, they require access to hierarchical relations between concepts (dog isAn animal) as well as non-hierarchical ones (gasoline fuels car). Knowledge resources such as lexical databases, semantic networks and ontologies explicitly encode such conceptual knowledge. However, traditionally, these have been manually created, which is expensive and time consuming for large re- sources, and cannot provide adequate coverage in specialised domains. In order to alleviate this acquisition bottleneck, statistical methods have been created to acquire lexical and conceptual knowledge automatically from text. In particular, unsupervised techniques have the advantage that they can be easily adapted to any domain, given some corpus on the topic. However, due to sparseness issues, they often require very large corpora to achieve high quality results. The spectrum of resources and statistical methods has a crucial gap in situations when manually cre- ated resources do not provide the necessary coverage and only limited corpora are available. This is the case for real-world domain applications such as an NLP system for processing technical information based on a limited amount of company documentation. We provide a large-scale demonstration that this gap can be filled through the use of automatically generated networks. The corpus is automatically transformed into a network representing the terms or concepts which occur in the text and their relations, based entirely on linguistic tools. The net- works structurally lie in between the unstructured corpus and the highly structured manually created resources. We show that they can be useful in situations for which neither existing approach is ap- plicable. In contrast to manually created resources, our networks can be generated quickly and on demand. Conversely, they make it possible to achieve higher quality representations from less text than corpus-based methods, relieving the requirement of very large scale corpora. We devise scaleable frameworks for building networks from plain text and Wikipedia with varying levels of expressiveness. This work creates concrete networks from the entire British National Corpus covering 1.2m terms and 21m relations and a Wikipedia network covering 2.7m concepts. We develop a network-based semantic space model and evaluate it on the task of measuring semantic relatedness. In addition, noun compound paraphrasing is tackled to demonstrate the quality of the indirect paths in the network for concept relation description. On both evaluations we achieve results competitive to the state of the art. In particular, our network-based methods outperform corpus-based methods, demonstrating the gain created by leveraging the network structure.

The uncontrolled usage of hashtags in social media makes them vary a lot in the quality of semantics and the frequency of usage. Such variations pose a challenge to the current approaches which capitalize on either the lexical semantics of a hashtag by using metadata or the contextual semantics of a hashtag by using the texts associated with a hashtag. This thesis presents a hybrid approach to clustering hashtags based on their semantics, designed in two phases. The first phase is a sense-level metadata-based semantic clustering algorithm that has the ability to differentiate among distinct senses of a hashtag as opposed to the hashtag word itself. The gold standard test demonstrates that sense-level clusters are significantly more accurate than word-level clusters. The second phase is a hybrid semantic clustering algorithm using a consensus clustering approach which finds the consensus between metadata-based sense-level semantic clusters and text-based semantic clusters. The gold standard test shows that the hybrid algorithm outperforms both the text-based algorithm and the metadata-based algorithm for a majority of ground truths tested and that it never underperforms both baseline algorithms. In addition, a larger-scale performance study, conducted with a focus on disagreements in cluster assignments between algorithms, shows that the hybrid algorithm makes the correct cluster assignment in a majority of disagreement cases.

The number of R extension packages available from the CRAN repository has tremendously grown over the past 10 years. We look at this phenomenon in more detail, and discuss some of its consequences. In particular, we argue that the statistical computing community needs a more common understanding of software quality, and better domain-specific semantic resources.

The rapid development of mobile technologies has facilitated users to generate and store files on mobile devices. However, it has become a challenging issue for users to search efficiently and effectively for files of interest in a mobile environment that involves a large number of mobile nodes. In this thesis, file management and retrieval alternatives have been investigated to propose a feasible framework that can be employed on resource-limited devices without altering their operating systems. The file annotation and retrieval framework (FARM) proposed in the thesis automatically annotates the files with their basic file attributes by extracting them from the underlying operating system of the device. The framework is implemented in the JME platform as a case study. This framework provides a variety of features for managing the metadata and file search features on the device itself and on other devices in a networked environment. FARM not only automates the file-search process but also provides accurate results as demonstrated by the experimental analysis. In order to facilitate a file search and take advantage of the Semantic Web Technologies, the SemFARM framework is proposed which utilizes the knowledge of a generic ontology. The generic ontology defines the most common keywords that can be used as the metadata of stored files. This provides semantic-based file search capabilities on low-end devices where the search keywords are enriched with additional knowledge extracted from the defined ontology. The existing frameworks annotate image files only, while SemFARM can be used to annotate all types of files. Semantic heterogeneity is a challenging issue and necessitates extensive research to accomplish the aim of a semantic web. For this reason, significant research efforts have been made in recent years by proposing an enormous number of ontology alignment systems to deal with ontology heterogeneities. In the process of aligning different ontologies, it is essential to encompass their semantic, structural or any system-specific measures in mapping decisions to produce more accurate alignments. The proposed solution, in this thesis, for ontology alignment presents a structural matcher, which computes the similarity between the super-classes, sub-classes and properties of two entities from different ontologies that require aligning. The proposed alignment system (OARS) uses Rough Sets to aggregate the results obtained from various matchers in order to deal with uncertainties during the mapping process of entities. The OARS uses a combinational approach by using a string-based and linguistic-based matcher, in addition to structural-matcher for computing the overall similarity between two entities. The performance of the OARS is evaluated in comparison with existing state of the art alignment systems in terms of precision and recall. The performance tests are performed by using benchmark ontologies and the results show significant improvements, specifically in terms of recall on all groups of test ontologies. There is no such existing framework, which can use alignments for file search on mobile devices. The ontology alignment paradigm is integrated in the SemFARM to further enhance the file search features of the framework as it utilises the knowledge of more than one ontology in order to perform a search query. The experimental evaluations show that it performs better in terms of precision and recall where more than one ontology is available when searching for a required file.

Learning and representing semantics is one of the most important tasks that significantly contribute to some growing areas, as successful stories in the recent survey of Turney and Pantel (2010). In this thesis, we present an in- novative (and first) framework for creating a multimodal distributional semantic model from state of the art text-and image-based semantic models. We evaluate this multimodal semantic model on simulating similarity judgements, concept clustering and the newly introduced BLESS benchmark. We also propose an effective algorithm, namely Parameter Estimation, to integrate text- and image- based features in order to have a robust multimodal system. By experiments, we show that our technique is very promising. Across all experiments, our best multimodal model claims the first position. By relatively comparing with other text-based models, we are justified to affirm that our model can stay in the top line with other state of the art models. We explore various types of visual features including SIFT and other color SIFT channels in order to have prelim- inary insights about how computer-vision techniques should be applied in the natural language processing domain. Importantly, in this thesis, we show evi- dences that adding visual features (as the perceptual information coming from...

In the present study, we conducted an event-related potentials (ERP) study to examine episodic and semantic memory. We focused on two well-known patterns: the semantic N400 and the old/new fN400. Some researchers have argued that they reflect the same neuropsychological response (Voss & Federmeier, 2011). Others have suggested that they have distinct spatial-temporal signatures and reflect different psychological processes (Bridger, Bader, Kriukova, Unger, & Mecklinger, 2012). In the present study, we analyzed data using the basic N400/fN400 paradigm. We expect to find similar results to Bridger et al. (2012) in that the N400 and fN400 to be reliably different in topography and function.

Accurate situational assessment is key to any decision maker and especially crucial in military command and control, air traffic control, and complex system decision making. Endsley described three dependent levels of situational awareness, (1) perception, (2) understanding, and (3) projection. This research was focused on Endsley's second-level situational awareness (understanding) as it applies to service-oriented information technology environments in the context of the Semantic Web. Specifically, this research addressed the problem of developing accurate situational assessments related to the status or health of information technology (IT) services, especially composite, dynamic IT services, when some of Endsley's first level (perceived) information was inaccurate or incomplete. Research had not adequately addressed the problem of how to work with inaccuracy and situational awareness information in order to produce accurate situational assessments for Semantic Web services. This problem becomes especially important as the current Web moves towards a Semantic Web where information technology is expected to be represented and processed by machines. Costa's probabilistic Web ontology language (PR-OWL), as extended by Carvalho (PR-OWL2), is a framework for storage of and reasoning with uncertainty information as part of the Semantic Web. This study used Costa's PR-OWL framework, as extended by Carvalho, to build an ontology that supports reasoning with service-oriented information in the context of the Semantic Web and then assessed the effectiveness of the developed ontology through the use of competency questions, as described by Gruninger and Fox and verified through the use of an automated reasoner. This research resulted in a Web Ontology Language for Services (OWL-S), PR-OWL2 based ontology, and its associated Multi-Entity Bayesian Network which are flexible and highly effective in calculating situational assessments through the propagation of posterior probabilities using Bayesian logic. Specifically, this research (1) identifies sufficient information required for effective situational awareness reasoning, (2) specifies the predicates and semantics necessary to represent service components and dependencies, (3) applies Multi-Entity Bayesian Network to reason with situational awareness information, (4) ensures the correctness and consistency of the situational awareness ontology, and (5) accurately estimates posterior probabilities consistent with situational awareness information.