Faculty of Computer Science (Fredericton)
Pages
-
-
Deep belief networks for sentiment analysis
-
by Yong Jin, Sentiment analysis is a highly popular issue both in academic and engineering fields. Nowadays there is an increasingly large amount of online opinion resources, so people are inclined to develop some systems that can automatically determine the polarities of opinions, especially in the decision-making process of a company. On the other hand, deep learning is a recently developed popular topic and has received much attention in machine learning area. Deep belief network (DBN) is one important deep learning model, which has proved powerful in many domains including natural language processing. However, there still exist some big challenges for DBNs in sentiment analysis because of the complexity to express opinions. Therefore, this study tries to improve DBNs in sentiment analysis area from the following three aspects: (1) The neuron models are investigated in DBNs for sentiment prediction. We perform various experiments and apply both total accuracy and F-measure to evaluate the performance, which proves that Gaussian neuron model with specific parameter setting has better efiect. (2) In addition to the traditional bag-of-words representation for each sentence, the word positional information is considered in the input. We propose a new word positional contribution form and another novel word-to-segment matrix representation for text to incorporate the positional information into DBNs for sentiment analysis. Finally, we evaluate the performance via the total accuracy. The results show that the word positional information of sentences helps to improve the performance of DBNs for sentiment classification. (3) We propose a new method to improve the DBN learning algorithm based on the unsupervised training phase of restricted Boltzmann machines (RBMs). That is, the RBM generates the hidden layer in an unsupervised fashion, and then we use this hidden layer as the output of a single-layer neural network, which is trained via the delta rule (DR). The new weights trained from DR are then transmitted into the whole network for initialization of back propagation (BP). This way keeps more correction signal for each layer in the BP algorithm compared to the ordinary DBN training. Our experimental results demonstrate that this updated learning method performs better than the ordinary learning for sentiment classification., Ph.D. University of New Brunswick, Faculty of Computer Science, 2017.
-
-
Design and implementation of a distributed rule-based query system supporting conference organization
-
by Chaudhry Usman Ali, Conference organization involves a multitude of procedures consuming much time and effort of their Organization Committees ( OCs). Conference organization systems attempt to alleviate the burden of repetitive tasks through the (partial) automation of organizational processes. This thesis is focused on the design and implementation of automated query answering about a conference, retrieving and deriving QC-related information for use by (other) OC members, (candidate) PC members, (prospective) authors, as well as (potential) partners, sponsors, and participants. The Rule Responder framework is instantiated to a distributed rule-based system relieving OC members from answering such routine requests. Each team of co-chairs from the symposium's OC is supported by a Personal Agent (PA) that uses a local knowledge base containing co-chair facts and rules to answer queries for which the co-chairs are responsible. The External Agent (EA) acts as a single point of entry for users to interact with the system employing a Web form coupled to an HTTP port to which post and get requests are sent. The system has three Organizational Agents ( OAs), where one Super-Organizational Agent (Super-OA) acts as a dispatching manager to direct requests sent by a user via the EA to one of the two Sub-Organizational Agents (Sub-OAs): The “Event” Sub-OA deals with queries about the (‘temporary’) conference edition while the “Structure” Sub-OA handles queries about the (‘permanent’) institution holding the conference series. These Sub-OAs further delegate the requests to underlying PAs representing local knowledge of, respectively, the conference's (temporary) OC co-chairs and the institution's (permanent) subgroup co-chairs. The designed query-answering architecture has been implemented, evaluated, and deployed in the SymposiumPlanner-2012 use case supporting the RuleML-2012 Symposium. General design principles and implementation techniques for future conference planners are distilled from the lessons learnt from this use case., Scanned from archival print submission., M.C.S. University of New Brunswick, Faculty of Computer Science, 2013.
-
-
Design and implementation of peer collaboration service framework on cloud
-
by Dong Dong, Most of the key tasks or work in today's business are strongly related to collaboration. One of the important reasons that people collaborate is to complete a task which is hard to be done by individuals independently. With the prevalence of the Internet and mobile devices accessing the Internet with high-bandwidth network, it is easier for people in different locations to form groups anywhere and anytime. However, there are few methods to manage these dynamic web based collaborations. This thesis describes implementation of a framework named \Peer Collaboration Service Framework" providing a systematic approach to create and manage network based dynamic peer collaborations. The framework consists of three layers: (1) collaboration as a service layer, consisting of services to generate peer collaborations; (2) collaboration service layer, consisting of services running at the back end of collaborations to support them; (3) collaboration instance layer, supporting the generated collaboration application instances used by participants. This framework is implemented on Amazon EC2 cloud computing platform and employs several other web services offered by Amazon. A case study on collaborative software testing applications and experiments are also presented in the thesis., Electronic Only.
(UNB thesis number) Thesis 9199.
(OCoLC) 960908841, M.C.S., University of New Brunswick, Faculty of Computer Science, 2013.
-
-
Design pattern as a service for service-oriented systems
-
by Eltaher Mohamed El-Shanta, Software systems nowadays face many more challenges than ever before due to the intrinsic high complexity of systems and increasing demands from organizations. While patterns enable reuse of abstract design and architectural knowledge, abstractions documented as patterns do not directly yield reusable code. Software design patterns in particular are essential building blocks for almost any software system. In an effort to enable the use and reuse of implemented software design patterns, we propose a methodology to implement software design patterns as pattern services to make building pattern-based software applications considerably easier and faster. We describe the conceptual architecture and steps of the proposed methodology, and then we explain the implementation stages of a pattern as a service. After that, we demonstrate how the proposed methodology can be applied to Service-Oriented Architecture (SOA) patterns. To create a platform for managing pattern services, we design a Pattern as a Service (PaaS) system that functions as the platform for developing, storing, integrating, deploying, and managing pattern services and pattern-based applications. Furthermore, we describe a prototypical implementation of the PaaS system and the implementation of two case study applications, namely, an Online Discussion Group (ODG) and Online Stock Market Ticker (OSMT) that make use of the Observer pattern service and use the prototypical PaaS system as their platform. Then we perform some evaluation procedures on the proposed methodology both analytically and experimentally, and we give some concrete test results. Finally, we attach an appendix to this thesis in which we apply the methodology to the 23 design patterns introduced by Gamma et al. (1995). In it, we describe the important contents of each resulting pattern service.
-
-
Designing and realizing the USB interface in the thermal conductivity instrument
-
by Ning Ju, The Thermal Conductivity Instrument (TCi) is a state-of-the-art scientific instrument
for analyzing thermo-physical properties of materials. The product includes
multiple development elements, mechanical design, electronics hardware and firmware
design, software application on the PC side and scientific aspects in thermal conductivity
test. In our research we provide a way to expand the Universal Serial Bus
(USB) interface function for TCi inside which the microprocessor has no USB module.
Various approaches are considered including: Device firmware, USB firmware,
USB driver ( windows OS side) and Device application software. To be specific, what
is discussed and described in this thesis is a whole process about how to develop a
USB support product which means our solution can be applied to many products
rather than just the TCi.
-
-
Determining if this word is used like that word
-
by Milton King, Determining the meaning of a word in context is an important task for a variety of natural language processing applications such as translating between languages, summarizing paragraphs, and phrase completion. Usage similarity (USim) is an approach to describe the meaning of a word in context that does not rely on a sense inventory -- a set of dictionary-like definitions. Instead, pairs of usages of a target word are rated in terms of their similarity on a scale. In this thesis, we evaluate unsupervised approaches to USim based on embeddings for words, contexts, and sentences, and achieve state-of-the-art results over two USim datasets. We further consider supervised approaches to USim, and find that they can increase the performance of our models. We look into a more detailed evaluation, observing the performance on different parts-of-speech as well as the change in performance when using different features. Our models also do competitively well in two word sense induction tasks, which involve clustering instances of a word based on the meaning of the word in context., M.C.S. University of New Brunswick, Faculty of Computer Science, 2017.
-
-
Diachronically like-minded user community detection
-
by Hossein Fani, Diachronically like-minded user community detectionStudy of users’ behaviour, interests, and influence is of interest within the realm of online social networks due to its wide range of applications, such as personalized recommendations and marketing campaigns. However, the proposed approaches are not always scalable to a large number of users and a huge amount of user-generated content. Community-level studies are introduced to facilitate scalability, among other characteristics, highlighting the main properties of the network at a higher collective level. Prior work is mainly focused on the identification of online communities that are formed based on shared links and/or similar content. However, there is little literature on detecting communities that simultaneously share topical and temporal similarities. To extract diachronically like-minded user communities who have similar temporal dispositions according to their topics of interest from social content, we put forward two approaches: i) multivariate time series analysis, and ii) neural embeddings. In the former approach, we model users’ temporal topics of interest through multivariate time series, and inter-user affinities are calculated based on pairwise cross-correlation. While simple and effective, this approach suffers from sparsity in multivariate time series. In the latter method, however, each user is mapped to a dense embedding space and inter-user affinities are calculated based on pairwise cosine similarity.
While the objective of these two proposed approaches is to identify user communities up until the present; in the last step of this thesis, we propose two approaches to identify future communities, i.e., community prediction: i) Granger regression, and ii) temporal latent space modeling. In Granger regression, we propose to consider both the temporal evolution of users’ interests as well as inter-user influence through the notion of causal dependency. In the latter method, however, we assume that each user lies in an unobserved latent space, and similar users in the latent space are more likely to be members of the same user community. The model allows each user to adjust her location in the latent space as her topics of interest evolve over time.
Empirically, we demonstrate that our proposed approaches, when evaluated on a Twitter dataset, outperform existing methods under two application scenarios, namely news recommendation and user prediction., Electronic Only.
-
-
Diskless data analytics on distributed coordination systems
-
by Dayal Dilli, A distributed system contains software programs, applications and data resources dispersed across independent computers connected through a communication network. Distributed coordination systems are file-system like distributed meta-data stores that ensure consistency between processes of the distributed system. The challenge in this area is to perform processing fast enough on data that is continuously changing. The focus of this research is to reduce the disk bound time of a chosen distributed coordination system called Apache Zookeeper. By reducing the disk dependency, the performance will be improved. The shortcoming of this approach is that data is volatile on failures. The durability of the data is provided by replicating the data and restoring it from other nodes in the distributed ensemble. On average, a 30 times write performance improvement has been achieved with this approach.
-
-
Diskless data analytics on distributed coordination systems
-
by Dayal Dilli, A distributed system contains software programs, applications and data resources dispersed across independent computers connected through a communication network. Distributed coordination systems are file-system like distributed meta-data stores that ensure consistency between processes of the distributed system. The challenge in this area is to perform processing fast enough on data that is continuously changing. The focus of this research is to reduce the disk bound time of a chosen distributed coordination system called Apache Zookeeper. By reducing the disk dependency, the performance will be improved. The shortcoming of this approach is that data is volatile on failures. The durability of the data is provided by replicating the data and restoring it from other nodes in the distributed ensemble. On average, a 30 times write performance improvement has been achieved with this approach., Electronic Only.
(UNB thesis number) Thesis 9336.
(OCoLC) 961212958, M.C.S., University of New Brunswick, Faculty of Computer Science, 2014.
-
-
Distributed modular ontology reasoning
-
by Li Ji, This thesis proposes algorithms for a distributed reasoning system over
interface-based modular ontologies. The thesis research includes three
parts: (1) The algorithm designs for distributed modular ontology
reasoning, including TBox (see Glossary 5.) and ABox (see Glossary 1.)
reasoning of concept, negated concept, disjunction, conjunction,
subsumption, and role queries; (2) The distributed modular ontology
reasoning system, comprising system functionality, architecture and
functionality realization; and (3) A case study and experiments for
evaluating the distributed modular ontology reasoning compared to
monolithic ontology reasoning.
Pages
Zircon - This is a contributing Drupal Theme
Design by
WeebPal.
