https://www.um.edu.mt/library/oar/handle/123456789/62700 2026-04-14T23:42:58Z https://www.um.edu.mt/library/oar/handle/123456789/63048 Title: Investigating toxicity in 'Overwatch' Abstract: This study is centered on the investigation of how machine learning models would perform in automatically detecting toxicity, how much their performance would improve and the measuring of how much toxicity is present in the game chosen. The models that performed the best with the data were SVM and Decision Tree and both reached an accuracy of instances classified correctly of 81.9231%. It was found that the acoustic features used in this study (pitch and intensity) helped with the performance of the models trained to automatically detect toxicity. Although the study presents interesting results and shows that the task of detecting toxicity automatically in general is a feasible one, it is based on a relatively limited sample, such that these results cannot be considered as a representation of the overall population. Lastly, due to the changes in Twitch.tv private policy, streamers were being sanctioned for being “toxic” and some of them got banned from this platform at first, and then started moderating their behavior on their return, which is believed to have affected the efficient gathering of the data. Description: B.SC.(HONS)HUMAN LANGUAGE TECH. 2020-01-01T00:00:00Z https://www.um.edu.mt/library/oar/handle/123456789/63045 Title: Intelligent speech recognition data acquisition for Maltese Abstract: Automatic Speech Recognition is a difficult task for under-resourced languages such as Maltese, as large quantities of data are required for its development. This dissertation seeks to provide a solution to this issue by crowdsourcing speech recordings and devising ways of validating this data efficiently. Common Voice was used as a crowdsourcing platform, facilitating the collection of 11+hours of speech data since its launch for Maltese. For validation, phonological analysis was performed on the text prompts using a grapheme-to-phoneme tool. The results of this were then compared to the number of syllables and segments detected in the speech using syllable nucleus detection and unsupervised automatic phoneme segmentation. Syllable distance between recordings and prompts was seen to be an effective metric for validation down to distances as small as a single syllable. Segment distance was effective when faced with differences of a few syllables or more. Description: B.SC.(HONS)HUMAN LANGUAGE TECH. 2020-01-01T00:00:00Z https://www.um.edu.mt/library/oar/handle/123456789/63040 Title: Teach me how to feel : learning to classify emotions from short texts Abstract: Emotion classification is a very useful process in natural language processing. It provides readers with more pragmatic information about the text that they are reading, but due to the lack of multimodal information, this can be quite hard todo computationally. The SemEval shared task of 2018, called Affect in Tweets, provided users withtweets with labels indicating which emotions they expressed. Eleven different emotions were used as classes, namely Anger, Anticipation, Disgust, Fear, Joy, Love,Optimism, Pessimism, Sadness, Surprise and Trust. In this dissertation, four different machine learning algorithms were tested to try and find the best approach for emotion classification in tweets. These were Naive Bayes, Logistic Regression, Support Vector Machines and a Recurrent Neural Network. Moreover, for the NaiveBayes and deep learning model, two different types of classes were used. Firstly, all eleven emotions were concatenated together and a binary string was created, representing the combination of all the emotions. Secondly, they were classified separately as individual classes and the model was trained on each emotion on its own. For the other two machine learning algorithms, only the latter strategy was used. Additionally, seven types of features and combinations were extracted from the tweets. These were: The tweets with a pre-processing procedure applied to them ; The subjectivity of the tweets ; The polarity of the tweets ; The tweets and their subjectivity ; The tweets and their polarity ; The values of subjectivity and polarity of the tweets ; The tweets and their subjectivity and polarity concatenated together. Upon evaluation, it turned out that the Logistic Regression model with the combination of Tweets, Subjectivity and Polarity used as features was the one that performed best. Furthermore, the correlation between emotions was analysed and it was found out that Anger and Disgust were the two most correlated emotions. The results obtained were carefully analysed, and conclusions were drawn about the feasibility of classifying emotions from short social media texts. Description: B.SC.(HONS)HUMAN LANGUAGE TECH. 2020-01-01T00:00:00Z https://www.um.edu.mt/library/oar/handle/123456789/63037 Title: Investigating the relationship between creativity and lexical networks Abstract: The aim of this dissertation is to study the relationship between creativity and mental lexicons, through the use of an alternative uses test, a lexical association task and word embedding models. The hypothesis was that there is a positive correlation between the score that an individual obtains on a standardised, alternative uses creativity test and the semantic distance between associations they mention in a targeted lexical association task (the semantic distance being taken to be synonymous to inverse cosine similarity measure of the vectors of the words in a word embedding model). Each of 18 participants took part in an alternative uses test which was followed by a lexical association task. Responses collected were then analysed by the author of this dissertation. For the alternative uses creativity test, the score obtained by the participant was manually computed, based on a pre-devised marking scheme. For the lexical associations, the semantic distance between concepts was calculated based on the inverse similarity measures of the vectors, in two pre-trained word embedding models (Word2Vec and GloVe). The results for both tasks were compared and Pearson’s correlations were calculated between the normalised angular distance between words mentioned in the lexical association task, for both the total creativity scores, as well as for the four components of creativity (Fluency, Elaboration, Flexibility, Originality). A significant correlation was found between creativity and the semantic distances of words given in a lexical associations task. The fluency component of creativity was the most correlated creativity component to semantic distance. The order in which associations for the lexical association task were given did not make a difference to the size of the correlation between the creativity and lexical association task scores. Description: B.SC.(HONS)HUMAN LANGUAGE TECH. 2020-01-01T00:00:00Z