The recent surge in the field of generative artificial intelligence (GenAI) has the potential to bring about transformative changes across a range of sectors, including software engineering and education. As GenAI tools, such as OpenAI's ChatGPT, are increasingly utilised in software engineering, it becomes imperative to understand the impact of these technologies on the software product. This study employs a methodological approach, comprising web scraping and data mining from LeetCode, with the objective of comparing the software quality of Python programs produced by LeetCode users with that generated by GPT-4o. In order to gain insight into these matters, this study addresses the question whether GPT-4o produces software of superior quality to that produced by humans. The findings indicate that GPT-4o does not present a considerable impediment to code quality, understandability, or runtime when generating code on a limited scale. Indeed, the generated code even exhibits significantly better values across all the three code quality dimensions in comparison to the user-written code. However, no significantly superior values were observed for the generated code in terms of memory usage in comparison to the user code, which contravened the expectations. Furthermore, it will be demonstrated that GPT-4o encountered challenges in generalising to problems that were not included in the training data set. This contribution presents a first large-scale study comparing generated code with human-written code based on LeetCode platform based on multiple measures including code quality, code understandability, time behaviour and resource utilisation. All data is publicly available for further research.

In light of the imminent labor shortage, companies find themselves compelled to enhance their standing as desirable employers in the competitive pursuit of talent. This endeavor can be facilitated through the medium of online job advertisements (OJAs), which serve as a means to subtly communicate the merits of an organization to prospective employees. The objective of this study is to develop and evaluate a topic-modeling approach, known as Latent Dirichlet Allocation (LDA), for analyzing online data. The study will also discuss the strengths and weaknesses of the approach. The work will also include considerations on social sciences theory, signaling theory, and methods will be evaluated.

The automated classification of job titles constitutes a critical component of labor market research, survey analysis, and administrative data processing. The present study explores the classification of German job titles according to the German Classification of Occupations (KldB), with a particular emphasis on the linguistic and structural challenges that are inherent to this task. This study builds upon previous research by incorporating a variety of heterogeneous data sources, including manually annotated survey responses, a comprehensive synonym dataset, online job advertisements (OJAs), and vocational education and training titles from DAZUBI. Conventional machine learning models, including logistic regression, naive Bayes, and random forest, are employed to assess the classification performance at varying taxonomic levels of the KldB. The findings of the present study demonstrate that while substantial results can be achieved for broad occupational categories, fine-grained classification, particularly at the level of performance (5th digit), remains challenging. The findings underscore the limitations of relying solely on job titles and underscore the importance of richer contextual information and more expressive models. This work provides both an expanded dataset and a systematic analysis of classification performance, thereby establishing the foundation for future research on context-aware occupational coding in the German labor market.

This habilitation thesis compiles research on the challenges of complex networks in com- puter science and their applications. It includes case studies on interdisciplinary research in life sciences, computational social sciences, and digital humanities. In the life sciences, knowledge graph approaches are commonly used for clinical and biomedical data. This thesis focuses on context mining, algorithmic challenges, and link prediction. In social sciences network approaches, the goal is to connect social network analysis with ontology- driven research on the labor market. Although data sets are frequently available in social sciences, this is not always the case in the humanities. Therefore, when applying complex network approaches such as social network analysis to textual data, hermeneutical and methodological considerations are necessary. Once these considerations are addressed, data science methods such as text mining can be used to construct networks from texts. This thesis presents two case studies on social network analysis, in addition to addressing the challenges of interdisciplinary research on complex networks in computer science. By describing three different domains, it demonstrates the existence of a common toolbox that utilizes methods from data science and graph theory. Consequently, this thesis argues for more interdisciplinary exchange.

In this paper, an approach is presented for semantically enriching and linking historical vocational education and training (VET) documents using an ontology-centric method grounded in occupations and educational programs. The present study draws on a digitized corpus of archival documents from various political regimes in Germany—including the German Empire, the German Democratic Republic (GDR), and the Federal Republic Germany (FRG) — in order to explore strategies for annotating job titles and aligning them with standardized taxonomies such as KldB and ISCO. The proposed methodology integrates phrase matching, classification models, and ontology-based linking via the German Labor Market Ontology (GLMO), thereby enabling cross-referencing of documents by occupation and educational structure. The proposed workflow is designed to support longitudinal studies and promote interoperability across fragmented archival collections. This offers a scalable solution for labor market and education research.

There are currently several approaches to managing longitudinal data in graphs and social networks. All of them influence the output of algorithms that analyse the data. We present an overview of limitations, possible solutions and open questions for different data schemas for temporal data in social networks, based on a generic RDF-inspired approach that is equivalent to existing approaches. While restricting the algorithms to a specific time point or layer does not affect the results, applying these approaches to a network with multiple time points requires either adapted algorithms or reinterpretation. Thus, with a generic definition of temporal networks as one graph, we will answer the question of how we can analyse longitudinal social networks with centrality measures. We present two approaches to approximate the change in degree and betweenness centrality measures over time with two new measures, “importance” and “change”, to identify nodes with specific behaviors and apply these in two examples from educational research describing longitudinal data in labor market related topics in social networks.

Valuable insights can be gained regarding jobs and professions across various sectors of society based on their inherent and acquired traits. Previous studies relied on methods such as action research, surveys, and questionnaires that are time-consuming and resource-intensive. This study examines vocational education and training data on Twitter. Although the data has been utilized in multiple studies, we will examine a vital research inquiry within computational social science: Is it plausible to employ Twitter/X data for analyzing vocational education and training in Germany or does the data display excessive bias? This investigation is infrequently explored since most researchers endeavor to discover representative samplings of larger subsets, and gauging representativeness against a ground truth can prove challenging. However, we will demonstrate that with research inquiry and statistical data, it is feasible to calculate a representative distance d, correction factors κ, and an overall bias γ. This provides a unique technique towards labor market research that makes novel data interoperable, which has not been considered in previous literature. Our approach is versatile and can be readily extended to other data.

Digital archives often describe their catalogs with digital objects where each record represents a physical document that can be found in the archive. While this is approach is most effective to describe data that is present in the archive, there are some archives where the absence of missing documents can impact its quality. To approach methods that can be used to assess the completeness of an archive, we use record linkage methods on the example of the occupations archive at the Federal Institute for Vocational Education and Training (BIBB), combined with the genealogy of vocational education. Our idea is that every record from the genealogy - which describes the history of vocational training in Germany - should have a matching record in the occupations archive - which contains the legal regulations that build the foundation for the genealogy - which implies a mapping between the two data sets. To this end, we create pairs of potential matches between records of both data sets and separate them in two classes: matches and non-matches. These pairs are then used to train different classifiers that should be able to categorize unseen record pairs. The results show that the the selected classifiers give reasonable results but need some more improvements for better reliability.

The skills and qualifications of IT professionals are constantly changing and under discussion. In particular, we see the impact of emerging technologies and tools, such as AI, on the labor market and their reflection in the broader scientific community, media and everyday life [Do19; He16]. Several approaches have discussed how to assess the impact of computer science education from the perspective of education and labor market research [HBE22; SHL22]. However, in order to uncover the complex dynamics surrounding vocational education, we need to take a closer look at the labor demand as well as the public perception and valuation of professionals. Therefore, we present a novel approach to work with social media data, in particular Twitter/X data. We have collected a large dataset of job-related X data (2007-2023) but see that a word list-based searching strategy is not enough. Most tweets are actually job offers and we will discuss several strategies to categorize tweets accordingly to make a meaning out if it. We focus on occupations in natural sciences, geography and informatics.