This thesis addresses the inherent complexity of discourse parsing in Natural Language Processing (NLP) by developing a multilingual framework implemented for Penn Discourse TreeBank (PDTB) datasets. Leveraging the advancements of Large Language Models (LLMs) and transformer architectures, the thesis proposes a hybrid methodology that integrates fine-tuned BERT models for Discourse Connective (DC) detection and argument span labeling with in-context learning strategies for Discourse Relation Recognition (DRR). The study bridges the gap between isolated sub-tasks and end-to-end processing by defining interconnected modules that link detection, labeling, and recognition phases. Evaluating this pipeline across seven datasets in English, Portuguese, and Turkish, the framework achieves performance on par with state-of-the-art models. Additionally, the thesis contributes a novel lightweight DC detection model and introduces a method to enhance implicit discourse relation recognition using machine translation techniques, demonstrating the efficacy of these approaches in both high- and low-resource linguistic contexts.

Date: 23.12.2025 / 15:00 Place: A-212

This thesis develops a framework for six degrees-of-freedom (6DoF) spatial audio capture and reproduction using higher-order Ambisonics and rigid spherical microphone arrays (SMAs). It introduces SPWIN (Sparse Plane-Wave Interpolation), a novel method for accurately reconstructing sound fields between microphone positions, enabling realistic, head-tracked VR experiences. The framework also presents a spherical harmonic–based method for estimating sound source directivity. Performance is validated through objective interaural coherence analysis and subjective MUSHRA tests in VR. These advancements close key gaps in spatial audio reproduction and align with emerging standards such as MPAI-CAE, supporting next-generation immersive media systems.

Date: 25.08.2025 / 13:30 Place: B-223

This thesis investigates multimodal data fusion and multicompartment image analysis in acute and chronic lung diseases. In a COVID-19 cohort, we integrated imaging, clinical, and viral genomic data, using sparse canonical correlation analysis and cooperative learning to explore inter-modality associations and predict intensive care unit admission. We leveraged Word2Vec to encode the viral genome. In an interstitial lung diseases cohort, we extracted lung and pulmonary artery radiomics features from chest computed tomography scans, demonstrating predictive value for pulmonary hypertension and transplant-free survival. We illustrated that multimodal data fusion and multicompartment image analysis mirror clinical decision-making processes and improve personalized prognostication.

Date: 16.07.2025 / 15:30 Place: A-108

The thesis examines reference selection in Turkish, focusing on how linguistic and cognitive factors shape the choice of referring expressions. A novel coreference corpus based on spontaneous, goal-directed dialogs was developed using a new annotation scheme. Statistical and machine learning analyses reveal that competition and distance metrics are the strongest predictors of referential form, while grammatical and speaker roles show weaker but significant effects. The findings support and extend key predictions from theories about reference production and comprehension by focusing on a typologically different language and using naturalistic data.

Date: 18.06.2025 / 10:00 Place: A-212

A common evolutionary origin of the Obscurin and Trio protein families was identified through phylogenetic analyses, indicating ancient domain shuffling events across vertebrate lineages. Building on these findings, the novel variant effect predictor TrioNsight was developed for TrioN-like DH domains. The predictor, which integrates structural, evolutionary and physicochemical features, achieved high accuracy in distinguishing pathogenic variants (MCC: 0.906; 1.0 indicates perfect prediction). This work bridges fundamental evolutionary biology and clinical application, offering both mechanistic insights into protein family evolution and a practical tool for the interpretation of disease-associated variants in functionally critical domains.

Date: 30.05.2025 / 10:00 Place: A-212

Flavor perception and gut-brain signaling influence eating behavior. The vagus nerve (VN) may become desensitized to high-calorie foods, promoting overeating, a process potentially reversible through VN stimulation. This study examined whether transcutaneous VNS (tVNS) modulates brain responses to food using EEG during chocolate milkshake consumption. Key findings: (1) Spontaneous eyeblink rates (a dopamine proxy) increased after food intake, suggesting dopaminergic engagement. (2) Event-related potentials (ERPs) time-locked to swallowing were only observed with food stimuli, validating the sip-and-swallow protocol for flavor research. (3) No significant tVNS effects were found on resting-state oscillations or ERPs, but the method captured food-specific neural responses. These results support the protocol’s utility for future gut-brain studies.

Date: 14.04.2025 / 14:00 Place: A-212

The thesis discusses optimizing alarm management systems in manufacturing. It introduces a novel data-driven method using the Tennessee Eastman Process to enhance alarm parameters, aiming to reduce missed critical alarms and improve process safety. Key contributions include associating disturbances with alarms, creating an alarm simulation platform, and optimizing alarm parameters. The study highlights the trade-off between alarm reaction delay and the number of alarms, emphasizing the importance of timely operator responses.

Date: 10.01.2025 / 14:45 Place: II-06

This thesis examines clozapine effects on cognitive function and brain connectivity in schizophrenia with the M1 muscarinic receptor polymorphism (rs2067477). Differences in cortical activity and connectivity between genotypes were assessed using an N-back working memory task and functional near infrared spectroscopy (fNIRS). Wild-type individuals exhibited higher cortical activation during the task, but had lower functional connectivity in the frontotemporal network compared to non-wild-types. The findings suggest different compensatory mechanisms by highlighting the genetic effects of clozapine-related neural responses and provide valuable information for personalized treatment approaches in schizophrenia.

Date: 10.01.2025 / 10:00 Place: A-212

Date: 20.12.2024 / 13:00 Place: A-212

Introducing the EnSCAN framework, we propose a pioneering algorithm to consolidate selected variants even across distinct platforms, thereby prioritizing candidate causative loci and enhancing ML outcomes through combining prior information captured from each multi-model of each dataset. The proposed ensemble algorithm utilizes chromosomal locations of SNVs by mapping to cytogenetic bands, along with the proximities between pairs and multi-model via Random Forest validations to prioritize SNVs and candidate causative genes for Alzheimer Disease. The scoring method is scalable and can be applied to any multi-platform genotyping study. We present how the proposed EnSCAN scoring algorithm prioritizes the candidate causative variants related to LOAD among three GWAS datasets.

Date: 06.09.2024 / 16:00 Place: B-116