We are currently working on the following research topics:
Natural language processing and text mining facilitate the targeted processing and analysis of large textual data sets. These technologies are currently employed in numerous research disciplines and business areas, largely due to recent advancements in technologies of generative AI. A variety of methods and algorithms are currently employed, drawing from related fields such as AI, data mining, and psychology.
The significant volume of physical and digital legal documents generated and exchanged within law firms, judicial authorities, and corporate legal departments presents a promising opportunity for the application of automatic text analysis approaches and techniques. This allows to uncover latent patterns, trends, and emotions that are not perceptible to humans, thereby enhancing the effectiveness of human decision-making processes.
In this research project we aim to examine and evaluate the utilization of diverse methodologies, techniques, and algorithms employed within legal contexts. We further investigate the potential benefits of these methods and the challenges that arise from their implementation for the design of corporate digital business processes.
In this research project we aim to investigate the current and future role of AI in the workplace and its implications for our work practices and collaboration with other individuals. Contemporary research questions are examined using empirical qualitative methods, with a particular focus on hybrid teams (consisting of humans and intelligent agents).
Considering legal work, the role of effective knowledge management to process legally relevant data is critical. Given that this field entails the processing of highly sensitive data, it raises significant concerns regarding which knowledge management systems are most suitable for storing codified knowledge. Therefore, we also address AI-based techniques for accessing collective legal knowledge, with the aim of preventing knowledge silos and facilitating intuitive knowledge acquisition across all relevant repositories.
The sheer volume of unstructured data being generated daily in both private and organizational environments, poses several challenges. For example, it's important to understand how to thoroughly analyze, assess and evaluate emotions in user-generated content without losing the context behind it.
In this research project, we explore among others the use of sentiment analysis and dictionaries as a proper solution approach frequently used in this field. In addition to creating and assessing domain-specific dictionaries to capture compound concepts, we study the role of emotions in various text styles and genres. Furthermore, we investigate how context affects the understanding of content created by users by developing and using classification schemes such as taxonomies and typologies.
Machine translation systems that are based on AI facilitate the automated translation of texts written in one language into one or more other languages. Various AI-based software solutions have long since found their way into our everyday private and professional life and already support us in accomplishing a variety of creative and knowledge-intensive tasks.
The results of these systems are typically evaluated based on technical metrics (such as the so-called BLEU score) or by researching user’s perception when accomplishing related translation tasks. Conversely, the function-related role perspective, which allows for the assessment of the actual impact of AI systems in the workplace, is frequently disregarded.
Thus, the objective of this research project is twofold: first, we investigate this role-related perspective in more detail based on empirical, quantitative research approaches, and second, we explore the potential of such translation services to support human task performers in coping with particularly knowledge-intensive and creative tasks. In this context, we also evaluate the linguistic quality of the translated results from the perspective of human perception. With our findings, we aim to provide a foundation for the subsequent formulation of recommendations concerning future collaboration between human and AI at work.
Large language models (LLMs) belong to the subfield of generative AI. They can independently generate new, coherent texts based on a large amount of training data. However, LLMs sometimes generate incorrect or fictitious answers, a phenomenon known as hallucination. Furthermore, the training data is often outdated and sometimes lacks industry specificity. Consequently, using LLMs in the legal field can have an impact on the quality of results. It can also affect the work with legal texts.
Self-hosted LLMs can be expanded with a Retrieval Augmented Generation (RAG) component, which allows the answers to be enriched with additional data from sources such as internal knowledge databases. This approach reduces the common issues of hallucinations and insufficient training data. This research project aims to evaluate the use of RAG in the legal domain using empirically methods.
