Microservices are increasingly gaining popularity in software design. It is essential for microservice architectures to have low response time variation to design testable and predictable systems. In this study, the aim is to predict the response time variation of microservice call graphs by using their topological features. Following the prediction processes with machine learning models, feature explanations methods are used to investigate which topological features are influential in the machine learning models' outputs regarding response time variation and how these features influence model outputs.

Date: 19.01.2024 / 09:00 Place: A-212

Compounding is one of the known ways of word formation. It is also a productive way of word-formation in German (Neef, 2009). Compounding in German makes use of some markers, mostly called linking elements, between the constituents, and this phenomenon is highly common. Whether these markers have any meaning or what primary functions they have are seemingly highly controversial. In this study, we suggest that the close relation between meaning and reference on the one hand and categorization on the other can be explored computationally in distributional properties of these markers which are difficult to identify analytically.

Date: 22.01.2024 / 14:00 Place: A-212

This study aims to investigate the potential outcomes when children are exposed to a hypothetical English, i.e. ergative English, rather than accusative English in the language acquisition process by using a child-directed speech data set. Based on the data set, the English grammar is constructed with both syntactic and semantic structures. Subsequently, some parts are modified for the hypothetical English. Following this, a model is trained to generate sentences with their corresponding syntactic and semantic structures. After the training process, a comparative analysis is conducted to determine the predominant category—accusative or ergative—in the acquisition of language by children.

Date: 22.01.2024 Place: A-212

This study investigates the global discourse organization of Turkish TEDx talks. 70 TEDx Talks in Turkish with, reliable human-generated transcriptions were chosen to be annotated. These were collected as subtitle files and manually annotated to map out significant discourse segments. For every talk, a number of features such as the number of total words, specific transition words, duration (second-wise), speed, average embedding and the ending percentile of each sentence were used for training Machine Learning (ML) models. The results indicate that the transitions between motivational discourse segments can be predicted with an F1-score of 0.78.

Date: 22.01.2024 / 13:00 Place: B-116

Translating natural language sentences into logical expressions has been a challenging task due to contextual information and the variational complexity of sentences. In recent years, a new deep learning architecture, namely the Transformer architecture, has been providing new ways to handle what was hard or seemed impossible in natural language processing tasks. The Transformer architecture and language models that are based on it revolutionized the artificial intelligence field of research and changed how we approach natural language processing tasks. In this thesis, we conduct experiments to see whether successful results can be achieved using Transformer models in translating sentences into first-order logic expressions.

Date: 23.01.2024 / 11:00 Place: B-116

In the United States, institutional investors submit 13F reports to SEC every quarter, disclosing the number of shares they own. However, investors may delay filing these reports to conceal their strategies. In this study, data from 13F filings were enriched with additional information such as stock prices and industries to create a bipartite, dynamic, and rich-attributed graph. The changes in network metrics, edge counts, and motif counts were visualized, highlighting crises in the relevant time period. The study found that clustering coefficient, significant position changes, late-filed 13F reports, and specific motif counts of graph shows significant changes during market crashes.

Date: 18.01.2024 / 10:00 Place: A-212

Cancers of unknown primary (CUP) pose treatment challenges due to unidentified primary sites, affecting survival rates significantly. Traditional methods falter in pinpointing origins. Gene expression-based studies offer promise; machine learning models trained on annotated cancer data achieve remarkable accuracy, surpassing conventional methods. Our study, utilizing three machine learning models trained on TCGA data, attained competitive accuracy, around 96%.

Date: 18.01.2024 / 11:00 Place: A-108

This thesis applies machine learning to predict outcomes of men’s singles tennis matches from 2009-2022, utilizing a standardized data mining framework, namely SRP-CRISP-DM, for replicable results. Employing six feature extraction techniques, three models, and two feature selection methods with time-based cross-validation and hyperparameter tuning, the Extreme Gradient Boosting model emerged as the top performer, scoring a Brier score of 0.1913 and an accuracy of 70.5%, with bookmakers' odds as the top predictive feature.

Date: 07.12.2023 Place: A-212

With the help of large language models, prompt engineering enables easy access to vast knowledge for various applications. However, limited research has been done on multi-hop question answering using this approach. This thesis introduces a new semi-automatic prompting method for answering two-hop questions. The method involves creating a prompt with automatically selected examples by grouping answer-named entities from the training set and using a chain-of-thought principle. The results demonstrate comparable performance to fine-tuned models on the MuSiQue dataset. Ablation studies further validate the effectiveness of each component in the proposed method. The approach has the potential to be applied to more complex multi-hop question-answering systems while upholding performance on par with other state-of-the-art techniques.

Date: 07.09.2023/13:00 Place: A-212

Crude oil prices are very volatile as they depend on many factors. Traders in this market, therefore need to constantly monitor different factors affecting the price, which can lead to information asymmetry for individual traders. Therefore, this thesis presents a comprehensive decision support system incorporating fundamental, technical, and sentiment analysis to support the decisions of crude oil traders.

Date: 06.09.2023 / 15:30 Place: A-212