Structure learning of context-specific graphical models

Abstract The ultimate problem considered in this thesis is modeling a high-dimensional joint distribution over a set of discrete variables. For this purpose, we consider classes of context-specific graphical models and the main emphasis is on learning the structure of such models from data. Traditional graphical models compactly represent a joint distribution through a factorization justi ed by statements of conditional independence which are encoded by a graph structure. Context-speci c independence is a natural generalization of conditional independence that only holds in a certain context, speci ed by the conditioning variables. We introduce context-speci c generalizations of both Bayesian networks and Markov networks by including statements of context-specific independence which can be encoded as a part of the model structures. For the purpose of learning context-speci c model structures from data, we derive score functions, based on results from Bayesian statistics, by which the plausibility of a structure is assessed. To identify high-scoring structures, we construct stochastic and deterministic search algorithms designed to exploit the structural decomposition of our score functions. Numerical experiments on synthetic and real-world data show that the increased exibility of context-specific structures can more accurately emulate the dependence structure among the variables and thereby improve the predictive accuracy of the models.

Learning English is Fun! Incidental vocabulary acquisition of music and traditional class ninth graders via listening to pop songs

Finnish youth are constantly exposed to music and lyrics in English in their free time. It is likely that this has a positive effect on vocabulary learning. Learning vocabulary while simultaneously accompanied with melodies is likely to result in better learning outcomes. The present thesis covers a study on the vocabulary learning of traditional and music class ninth graders in a south-western upper comprehensive school in Finland, mainly concentrating on vocabulary learning as a by-product of listening to pop music and learning vocabulary through semantic priming. The theoretical background presents viable linguistic arguments and theories, which provide clarity for why it would be possible to learn English vocabulary via listening to pop songs. There is conflicting evidence on the benefits of music on vocabulary learning, and this thesis sets out to shed light on the situation. Additionally, incorporating pop music in English classes could assist in decreasing the gap between real world English and school English. The thesis is a mixed method research study consisting of both quantitative and qualitative research materials. The methodology comprises vocabulary tests both before and after pop music samples and a background questionnaire filled by students. According to the results, all students reported liking listening to music and they clearly listened to English pop music the most. A statistically significant difference was found when analysing the results of the differences in pre- and post-vocabulary tests. However, the traditional class appeared to listen to mainstream pop music more than the students in the music class, and thus it seems likely that the traditional class benefited more from vocabulary learning occurring via listening to pop songs. In conclusion, it can be established that it is possible to learn English vocabulary via listening to pop songs and that students wish their English lectures would involve more music-related vocabulary exercises in the future. Thus, when it comes to school learning, pop songs should be utilised in vocabulary learning, which could also in turn result in more diverse learning and the students could, more easily than before, relate to the themes and topics of the lectures. Furthermore, with the help of pop songs it would be possible to decrease the gap between school English and real-world English.