This paper investigates the formality level of the language used in twenty articles from two sports that seem to cater to different social classes (soccer and horse polo). The articles that serve as the data were published in two different types of British newspapers, one broadsheet (The Daily Telegraph) and one tabloid (The Daily Express) from September 2010 through November 2017. The study uses a quantitative method by means of the F-measure, and a qualitative analysis of two articles whose results deviate from the rest. The quantitative results show that there is a difference in formality in sports articles on the two sports soccer and horse polo, where articles on polo score higher on the F-measure in both newspapers. Most articles on horse polo follow the pattern of the informational production with features like a high ratio of nouns, pronouns, long words, and adjectives often found in academic papers and legal documents etc. Articles on soccer follow the involved production, characterized by a high ratio of verbs, adverbs, pronouns, and WH-questions often found in spoken interaction. The qualitative analysis shows that the article on soccer which has a much higher F-score than the rest is an informative article on the price of season tickets, and that the polo article with a very low Fscore contained a lot of quoted speech. / I den här uppsatsen undersöks formalitetsnivån i tjugo artiklar om fotboll och hästpolo. Två sporter som vanligtvis har utövare från olika samhällsklasser. Artiklarna som använts som data har blivit publicerade i två olika typer av brittiska tidningar, en dagstidning (The Daily Telegraph) och en kvällstidning (The Daily Express) från september 2010 till november 2017. I studien används en kvantitativ metod kallad the F-measure och en kvalitativ analys av de två artiklar där resultaten skilde sig från övriga. De kvantitativa resultaten visar att det är skillnad på formaliteten i artiklarna om fotboll och hästpolo, där artiklar om hästpolo får ett högre Fvärde än artiklar om fotboll i båda tidningarna. Flertalet artiklar om hästpolo följer mönstret för informativa texter som karaktäriseras av ett högt antal substantiv, pronomen, adjektiv och långa ord av den typ som ofta finns i akademiska uppsatser och juridiska dokument etc. Artiklar om fotboll följer oftast mönstret för involverade texter, som kännetecknas av ett högt antal av verb, adverb, pronomen och frågeordsfrågor som ofta hittas i talat språk. Den kvalitativa analysen visar att fotbollsartikeln som hade ett mycket högre F-värde än övriga var en informativ artikel om priser på säsongsbiljetter, och att poloartikeln som hade ett väldigt lågt F-värde innehöll en hel del citat från intervjuer.

In this thesis, the performance of two over-sampling techniques, SMOTE and ADASYN, is compared. The comparison is done on three imbalanced data sets using three different classification models and evaluation metrics, while varying the way the data is pre-processed. The results show that both SMOTE and ADASYN improve the performance of the classifiers in most cases. It is also found that SVM in conjunction with SMOTE performs better than with ADASYN as the degree of class imbalance increases. Furthermore, both SMOTE and ADASYN increase the relative performance of the Random forest as the degree of class imbalance grows. However, no pre-processing method consistently outperforms the other in its contribution to better performance as the degree of class imbalance varies.

This thesis deals with image segmentation of volumetric medical data. It describes a well-known watershed technique that has received much attention in the field of medical image processing. An application for a direct segmentation of 3D data is proposed and further implemented by using ITK and VTK toolkits. Several kinds of pre-processing steps used before the watershed method are presented and evaluated. The obtained results are further compared against manually annotated datasets by means of the F-Measure and discussed.

This thesis involves parsing document-based reports from the United States Human Rights Reports and rating the human practices for various countries based on the CIRI (Cingranelli-Richards) Human Rights Data Project dataset. The United States Human Rights Reports are annual reports that cover internationally recognized human rights practices regarding individual, civil, political, and worker rights. Students, scholars, policymakers, and analysts used the CIRI data for practical and research purposes. CIRI analyzed the annual reports from 1981 to 2011 and then stopped releasing the dataset for any further years, but a possible reason is due to the manual process of scouring the Human Rights Reports and then rating each human rights practice for each country. This manual process provides a solid foundation for creating a new automated process. The automated process uses the rating values provided by CIRI in the 1981-2011 dataset as expected values to evaluate the accuracy of the rating process. To transition to an automated process, the General Architecture for Text Engineering (GATE) application is used. GATE is an open source project used for developing solutions for text processing. GATE is used in conjunction with the coding schemes provided within the CIRI Coding Manual to create an automated ratings process. The CIRI Coding Manual uses qualitative and quantitative criteria. The original and automated ratings are evaluated using GATE’s Annotation Diff Tool to get the F-measure for every country in the dataset. The evaluation cases range between 1999 and 2011 because those are the only years included in both the CIRI dataset and the Human Rights Reports. The F-measure results are more accurate when quantitative criteria is used to rate human rights practices. The primary contribution of this thesis is a method for automating each country’s human practice ratings so that the purpose of the CIRI project can be continued.

The aim of this thesis is comparison of selected classification methods which are logistic regression (binary and multinominal), multilayer perceptron and classification trees, CHAID and CRT. The first part is reminiscent of the theoretical basis of these methods and explains the nature of parameters of the models. The next section applies the above classification methods to the six data sets and then compares the outputs of these methods. Particular emphasis is placed on the discriminatory power rating models, which a separate chapter is devoted to. Rating discriminatory power of the model is based on the overall accuracy, F-measure and size of the area under the ROC curve. The benefit of this work is not only a comparison of selected classification methods based on statistical models evaluating discriminatory power, but also an overview of the strengths and weaknesses of each method.