<div>Politicians are increasingly using social media platforms, specifically the microblog Twitter, to interact with the public and express their stances on current policy issues. Due to this nearly one-on-one communication between politician and citizen, it is imperative to develop automatic tools for analyzing how politicians express their stances and frame issues in order to understand how they influence the public. Prior to my work, researchers have focused on supervised, linguistic-based approaches for the prediction of stance or agreement of the content of tweets and classification of the frames and moral foundations used to express a single tweet. The generalizability of these approaches, however, is limited by the need for direct supervision, dependency on current language, and lack of use of social and behavioral context available on Twitter. My works are among the first to study these general political strategies specifically for politicians on Twitter. This requires techniques capable of abstracting the textual content of multiple tweets in order to generalize across politicians, specific policy issues, and time. In this dissertation, I propose breaking from traditional linguistic baselines to leverage the rich social and behavioral features present in tweets and the Twitter network as a form of weak supervision for studying political discourse strategies on microblogs. My approach designs weakly supervised models for the identification, extraction, and modeling of the relevant linguistic, social, and behavioral patterns of Twitter. These models help shed light on the interconnection of ideological stances, framing strategies, and moral viewpoints which underlie the relationship between a politician's behavior on social media and in the real world. <br></div>

In this thesis we examine issues surrounding the development of software that predicts the function of biological sequences using sequence similarity. There is a pressing need for high throughput software that can annotate protein or DNA sequences with functional information due to the exponential growth in sequence data. In Chapter 1 we briefly introduce the molecular biology and bioinformatics that is assumed knowledge, and the objectives for the research presented here. In Chapter 2 we discuss the development of a method of comparing competing designs for software annotators, using precision and recall metrics, and a benchmark method referred to as Best BLAST. From this we conclude that data-mining approaches may be useful in the development of annotation algorithms, and that any new annotator should demonstrate its effectiveness against other approaches before being adopted. As any new annotator that utilises sequence similarity to predict the function of a sequence will rely on the quality of existing annotations, we examine the error rate of existing sequence annotations in Chapter 3. We develop a new method that allows for the estimation of annotation error rates. This involves adding annotation errors at known rates to a sample of reference sequence annotations that was found to be similar to query sequences. The precision at each error rate treatment is determined, and linear regression then used to find the error rate at estimated values for the maximum precision possible given assumptions concerning the impact of semantic variation on precision. We found that the error rate of curated annotations based on sequence similarity (ISS) is far higher than those that use other forms of evidence (49% versus 13-18%, respectively). As such we conclude that software annotators should avoid basing predictions on ISS annotations where possible. In Chapter 4 we detail the development of GOSLING, Gene Ontology Similarity Listing using Information Graphs, a software annotator with a design based on the principles discovered in previous chapters. Chapter 5 concludes the thesis by discussing the major findings from the research presented. / http://library.adelaide.edu.au/cgi-bin/Pwebrecon.cgi?BBID=1280882 / Thesis (M.Sc.(M&CS)) -- School of Computer Science, 2007

Diss. Univ. Bremen, 2007. / Computer Science (Springer-11645).

Master of Science / Department of Computing and Information Sciences / Doina Caragea / There are many formats in which digital information is stored in order to share and re-use it by different applications. The web can hardly be called old and already there is huge research going on to come up with better formats and strategies to share information. Ten years ago formats such as XML, CSV were the primary data interchange formats. And these formats were huge improvements over SGML (Standard Generalized Markup Language). It’s no secret that in last few years there has been a huge transformation in the world of data interchange. More lightweight, bandwidth-non-intensive JSON has taken over traditional formats such as XML and CSV. BigData is the next big thing in computer sciences and JSON has emerged as a key player in BigData database technologies. JSON is the preferred format for web-centric, “NoSQL” databases. These databases are intended to accommodate massive scalability and designed to store data which does not follow any columnar or relational model. Almost all modern programming languages support object oriented concepts, and most of the entity modeling is done in the form of objects. JSON stands for Java Script object notation and as the name suggests this object oriented nature helps modeling entities very naturally. And hence the exchange of data between the application logic and database is seamless. The aim of this report is to develop two similar applications, one with traditional SQL as the backend, and the other with a JSON supporting MongoDB. I am going to build real life functionalities and test the performance of various queries. I will also discuss other aspects of databases such as building a Full Text Index (FTI) and search optimization. Finally I will plot graphs to study the trend in execution time of insertion, deletion, joins and co- relational queries with and without indexes for SQL database, and compare them with the execution trend of MongoDB queries.

<p>In this thesis, we design frameworks for efficient and accurate floating point computation. The principle underlying our frameworks is making information usually discarded in the hardware, specifically, in the floating point unit, visible to the programmer. The programmer, or automated tools that we developed, can use this information to make accuracy/performance improvements or tradeoffs. </p><p>We make the error of floating point additions architecturally visible to programmers and experimentally demonstrate that programmers can use this error to improve the accuracy of their applications or improve the application's performance without affecting the accuracy of the final result. To free programmers from having to manually instrument their code, we develop a compiler pass to automate this process.</p><p>We also design a framework to profile applications to measure undesirable numerical behavior at the floating point operation level. We develop a debugger that programmers can use to find variables with "bad" behavior. In addition, we present a profile driven mixed precision analysis framework that heuristically determines the precision of all variables in an application based on their numerical behavior. We experimentally evaluate the mixed precision analysis to show that it can generate a range of results with different accuracies and precisions.</p> / Dissertation

<p> The Boolean Satisfiability Problem (SAT) is a canonical decision problem originally shown to be NP-complete in Cook's seminal work on the theory of computational complexity. The SAT problem is one of several computational tasks identified by researchers as core problems in computer science. The existence of an efficient decision procedure for SAT would imply P = NP. However, numerous algorithms and techniques for solving the SAT problem have been proposed in various forms in practical settings. Highly efficient solvers are now actively being used, either directly or as a core engine of a larger system, to solve real-world problems that arise from many application domains. These state-of-the-art solvers use the Davis-Putnam-Logemann-Loveland (DPLL) algorithm extended with Conflict-Driven Clause Learning (CDCL). Due to the practical importance of SAT, building a fast SAT solver can have a huge impact on current and prospective applications. The ultimate contribution of this thesis is improving the state of the art of CDCL by understanding and exploiting the empirical characteristics of how CDCL works on real-world problems. The first part of the thesis shows empirically that most of the unsatisfiable real-world problems solvable by CDCL have a refutation proof with near-constant width for the great portion of the proof. Based on this observation, the thesis provides an unconventional perspective that CDCL solvers can solve real-world problems very efficiently and often more efficiently just by maintaining a small set of certain classes of learned clauses. The next part of the thesis focuses on understanding the inherently different natures of satisfiable and unsatisfiable problems and their implications on the empirical workings of CDCL. We examine the varying degree of roles and effects of crucial elements of CDCL based on the satisfiability status of a problem. Ultimately, we propose effective techniques to exploit the new insights about the different natures of proving satisfiability and unsatisfiability to improve the state of the art of CDCL. In the last part of the thesis, we present a reference solver that incorporates all the techniques described in the thesis. The design of the presented solver emphasizes minimality in implementation while guaranteeing state-of-the-art performance. Several versions of the reference solver have demonstrated top-notch performance, earning several medals in the annual SAT competitive events. The minimal spirit of the reference solver shows that a simple CDCL framework alone can still be made competitive with state-of-the-art solvers that implement sophisticated techniques outside the CDCL framework.</p>

South Africans are confronted by social and economic problems of magnitude and complexity that cannot be ignored by curriculum planners. Increasing population statistics, plummeting gold prices, disinvestment, unemployment and political uncertainty continue to make economic and social ventures hazardous in their undertaking. Man accesses his world through education, and curriculum planners should, therefore, anticipate the future socio-economic structures of this country, and beyond, thereby providing meaningful school curricula that will prepare learners for the technologyintensive workplaces of modem times. Curriculating wisdom would include the: * needs of the learner as he prepares for adulthood; * demands of society for utilitarian citizens; and * structuring of knowledge that encourages learning rather than discourages it. The secondary school adolescent is career oriented and progresses towards a state of independent thought and behaviour. The school is obliged to address the situation of the adolescent and provide learning opportunities that will prepare him for a meaningful place in society. Selected themes from the discipline of computer science have been evaluated according to the essences of the pedagogic school structure. Differentiation of the subject content has been indicated so that pupils of all academic abilities can benefit from the learning experiences offered by the content. A questionnaire, sent to Education Departments in South Africa, determined that computer science was available to a very small percentage of pupils for matriculation creditation. A sample of these pupils completed a questionnaire designed to determine their regard for the subject and the influence it had on their choice of career. Three pupils were interviewed to assess, primarily, the role that computers played in their family and peer relationships. All investigations confirmed that computer science is a highly desired subject, skilfully able to fulfil the needs of the adolescent as he prepares for his place in society in the most meaningful, contemporaneous and dignified manner. / Educational Studies / D.Ed. (Philosophy of Education)

This thesis reports a study of network probing algorithms to wired and wireless Ethernet networks. H begins with a literature survey of Ethernet and related technology, and existing research on bandwidth probing. The OPtimized Network Engineering Tool (OPNET) was used to implement a network probing testbed, through the development of packet pair/stream modules. Its performance was validated using a baseline scenario (two workstations communicating directly on a wired or wireless channel) and it was shown how two different probe packet sizes allowed link parameters (bandwidth and the inter-packet gap) to be obtained from the packet pair measurements and compared with their known values. More tests were carried out using larger networks of nodes carrying cross-traffic, giving rise to multimodal dispersion distributions which could be automatically classified using data-clustering algorithms. Further studies used the ProbeSim simulation software, which allowed network and data classification processes were brought together in a common simulation framework The probe packet dispersion data were classified dynamically during operation, and a closed¬loop algorithm was used to adjust parameters for optimum measurement. The results were accurate for simple wired scenarios, but the technique was shown to be unsuitable for heterogeneous wired-cum-wireless topologies with mixed cross-traffic.

<p>Safety and efficiency are the most important factors in handling container cranes at ports all over the world. Rapid economic growth has led to a large increase of quay cranes in operation over the past decades, which is consequently paired with an increasing number of crane incidents. Crane operation becomes even more difficult with larger sized cranes, as the safety of these operations are solely depending on the experience of the operator. Thus, this heightens the demand for additional safety assistance devices. In this project, a camera based image processing design is introduced. By detecting the container that is being handled and adjacent ones at high speed, this system can predict and send a warning for a potential collision before the operator actually realizes the risk. </p><p> The proposed Edge Approaching Detection algorithm incorporated with the Hue, Saturation, and Value (HSV) algorithm is the key to this design. The combination of these two algorithms make it much faster to detect color-based objects at high speed and in real-time. By taking advantage of HSV&rsquo;s high efficiency, the computation required by traditional object detection is reduced dramatically. In this paper, this computation will be compared in terms of frames per second (FPS). As a result, accuracy is improved, speed is increased, and if possible, the switch to a cheaper platform powerful enough to support one specific project will reduce costs. </p>

The application of heuristic search techniques, such as genetic algorithms, to the problem of automatically generating software test data has been a growing interest for many researchers in recent years. The problem tackled by this thesis is the development of heuristics for test data search for a class of test data generation problems that could not be solved prior to the work done in this thesis because of a lack of an informative cost function. Prior to this thesis, work in applying search techniques to structural test data generation was largely limited to numeric test data and in particular, this left open the problem of generating string test data. Some potential string cost functions and corresponding search operators are presented in this thesis. For string equality, an adaptation of the binary Hamming distance is considered, together with two new string specific match cost functions. New cost functions for string ordering are also defined. For string equality, a version of the edit distance cost function with fine-grained costs based on the difference in character ordinal values was found to be the most effective in an empirical study. A second problem tackled in this thesis is the problem of generating test data for programs whose coverage criterion cost function is locally constant. This arises because the computation produced by many programs leads to a loss of information. The use of flag variables, for example, can lead to information loss. Consequently, conventional instrumentation added to a program receives constant or almost constant input and hence the search receives very little guidance and will often fail to find test data. The approach adopted in this thesis is to exploit the structure and behaviour of the computation from the input values to the test goal, the usual instrumentation point. The new technique depends on introducing program data-state scarcity as an additional search goal. The search is guided by a new fitness function made up of two parts, one depending on the branch distance of the test goal, the other depending on the diversity of the data-states produced during execution of the program under test. In addition to the program data-state, the program operations, in the form of the program-specific operations, can be used to aid the generation of test data. The program-specific operators is demonstrated for strings and an empirical investigation showed a fivefold increase in performance. This technique can also be generalised to other data types. An empirical investigation of the use of program-specific search operators combined with a data-state scarcity search for flag problems showed a threefold increase in performance.