• Our fine-tuned LLM handles resource-scarce scenarios better than previous SOTA approaches. • An instruction-fine-tuned LlaMA 3 8B achieves new SOTA performance for the ACSA and E2E tasks on the Rest-16 dataset and for the ACSA and TASD tasks for GERestaurant. • Few-shot prompting shows promising results, though fine-tuned LLMs typically achieve better results and are more efficient. • For fine-tuning LLMs, concise prompts are usually sufficient if combined with well-optimized hyperparameters. This study explores the use of fine-tuned open source large language models (LLMs) for Aspect-based Sentiment Analysis (ABSA), comparing their performance with state-of-the-art (SOTA) methods on English and German datasets with focus on low-resource scenarios. Results on the four ABSA subtasks Aspect Category Detection (ACD), Aspect Category Sentiment Analysis (ACSA), End-To-End-ABSA (E2E), and Target Aspect Sentiment Detection (TASD) show that fine-tuned LLMs handle limited training data scenarios better than current SOTA approaches, achieving consistent performance across various dataset sizes. Prompt formulation and hyperparameter tuning influence performance, though concise prompts often suffice when combined with effective fine-tuning. To assess generalizability, we conduct an ablation study across multiple languages, domains, and LLM architectures. The findings confirm that performance gains extend beyond the initial setting, supporting the robustness of fine-tuned LLMs over multiple different languages and domains. We establish new SOTA results on the Rest-16 and GERestaurant datasets and highlight the practical viability of fine-tuning LLMs for ABSA applications under limited training material.

Aspect-Based Sentiment Analysis (ABSA) is a fine-grained task in sentiment analysis, aiming to identify sentiment expressed towards specific aspects of an entity. This paper explores the use of Large Language Models (LLMs), specifically GPT-3.5-turbo and Llama-3-70B, for generating annotated data in Aspect-Based Sentiment Analysis (ABSA), aiming to address the scarcity of labelled datasets in the field. Two low-resource scenarios are considered, with 25 and 500 manually annotated examples available. In the 25-example scenario, adding synthetic examples generated through few-shot prompting resulted in F1 scores of 81.33 for Aspect Category Detection (ACD) and 71.71 for Aspect Category Sentiment Analysis (ACSA). For the 500-example scenario, synthetic data augmentation showed a notable gain only for the ACSA task, raising the F1 score from 84.54 to 86.70. • LLM-generated examples enhance performance in Aspect Category Detection (ACD). • Synthetic examples lead to an F1 score of 81.33 on the ACD task. • Llama-3-70B generated more linguistically diverse data than GPT-3.

Temporal regularities and the timing of events and actions such as anticipating enemy movements or planning one’s next move are essential components of almost every video game. Thus, to succeed in video games, it is advantageous to anticipate events and prepare relevant actions before they occur. This work explores whether elapsed time can be used as a predictive cue for implicitly anticipating events in video games. Inspired by findings from psychology, we implemented multiple time-event correlations in a custom video game by pairing specific delays with specific game events. Participants had to shoot targets that appeared at different locations. After a certain delay (e.g., 0.8 s), the targets appeared more frequently (80 % of all appearances) at a specific location (e.g., left up). Our analysis of 25 participants provides evidence that players implicitly learned the implemented time-event correlations and used them to anticipate the location of upcoming targets. This led to improved game performance. Although no participant realised the implemented temporal regularities, targets were shot faster when preceded by the frequently paired delay. Our findings pave the way for game developers and researchers alike to more creatively combine human temporal processing with temporal aspects of video games.

Feeling ownership of a virtual body is crucial for immersive experiences in VR. Knowledge about body ownership is mainly based on rubber hand illusion (RHI) experiments in the real world. Watching a rubber hand being stroked while one’s own hidden hand is synchronously stroked, humans experience the rubber hand as their own hand and underestimate the distance between the rubber hand and the real hand (proprioceptive drift). There is also evidence for a decrease in hand temperature. Although the RHI has been induced in VR, it is unknown whether effects in VR and the real world differ. We conducted a RHI experiment with 24 participants in the real world and in VR and found comparable effects in both environments. However, irrespective of the RHI, proprioceptive drift and temperature differences varied between settings. Our findings validate the utilization of the RHI in VR to increase our understanding of embodying virtual avatars.

Avatars in virtual reality (VR) enable embodied experiences and induce the Proteus effect—a shift in behavior and attitude to mimic one’s digital representation. Previous work found that avatars associated with physical strength can decrease users’ perceived exertion when performing physical tasks. However, it is unknown if an avatar’s appearance can also influence the user’s physiological response to exercises. Therefore, we conducted an experiment with 24 participants to investigate the effect of avatars’ athleticism on heart rate and perceived exertion while cycling in VR following a standardized protocol. We found that the avatars’ athleticism has a significant and systematic effect on users’ heart rate and perceived exertion. We discuss potential moderators such as body ownership and users’ level of fitness. Our work contributes to the emerging area of VR exercise systems.

In this paper, we present first work-in-progress annotation results of a project investigating computational methods of emotion analysis for historical German plays around 1800. We report on the development of an annotation scheme focussing on the annotation of emotions that are important from a literary studies perspective for this time span as well as on the annotation process we have developed. We annotate emotions expressed or attributed by characters of the plays in the written texts. The scheme consists of 13 hierarchically structured emotion concepts as well as the source (who experiences or attributes the emotion) and target (who or what is the emotion directed towards). We have conducted the annotation of five example plays of our corpus with two annotators per play and report on annotation distributions and agreement statistics. We were able to collect over 6,500 emotion annotations and identified a fair agreement for most concepts around a κ-value of 0.4. We discuss how we plan to improve annotator consistency and continue our work. The results also have implications for similar projects in the context of Digital Humanities.

We briefly describe a tutorial on the application of Eye-Tracking technology for Software Engineering Education. We will showcase our setup of a large-scale Eye-Tracking-Classroom and its usage for real-time improvement of traditional learning scenarios in Software Engineering Education. We will focus on the integration of gaze data into modern integrated development environments (IDEs) and demonstrate a complete workflow for its usage in co-located multi-user Eye-Tracking-Environments.

UML (Unified Modeling Language) is the current de facto as well as de jure standard (ISO/IEC 19505:2012) notation to visualize models in software development. UML provides essential guidelines and rules to visualize and understand complex software systems. This is the reason why it has become part of curricula for software engineering courses at many universities worldwide. It is well known, however, that UML is hard to grasp for novices, mainly due to its complexity. In order to tackle the problem of teaching UML to novice students appropriately, it is inevitable to understand their needs and problems much better than we do now. This paper presents empirical insights into students’ problems when developing common UML diagrams. Identified problems are generalized, giving rise to a problem catalogue that is derived from our empirical findings, thus establishing a basis for addressing these problems through focused learning arrangements.

Virtual reality (VR) allows embodying any possible avatar. Known as the Proteus effect, avatars can change users’ behavior and attitudes. Previous work found that embodying Albert Einstein can increase cognitive task performance. The behavioral confirmation paradigm, however, predicts that our behavior is also affected by others’ perception of us. Therefore, we investigated the cognitive performance in collaborative VR when self-perception and external perception of the own avatar differ. 32 male participants performed a Tower of London task in pairs. One participant embodied Einstein or a young adult while the other perceived the participant as Einstein or a young adult. We show that the perception by others affects cognitive performance. The Einstein avatar also decreased the perceived workload. Results imply that avatars’ appearance to both, the user and the others must be considered when designing for cognitively demanding tasks.

An exploratory literature review method was applied to publications from several sources on Human-Computer Interaction (HCI) for In-Vehicle Information Systems (IVIS). The novel approach for bibliographic classification uses a graph database to investigate connections between authors, papers, used methods, and investigated interface types. This allows the application of algorithms to find similarities between different publications and overlaps between different usability evaluation methods. Through community detection algorithms, the publications can be clustered based on similarity relationships. For the proposed approach several thousand papers were systematically filtered, classified, and stored in a graph database. The survey shows a trend for usability assessment methods with direct involvement of users, especially the observation of users and performance-related measurements, as well as questionnaires and interviews. However, especially methods usually applied in early stages of development based on the assessment through models or experts, as well as collaborative and creativity methods do not seem very popular in automotive HCI research.

This paper describes our implementation, teaching philosophy, and experiences with our C-based version of the widely known Karel the Robot introductory programming micro-language. Karel enables students to programmatically solve problems, using the C language, in a graphical two-dimensional world by moving the robot around while checking and manipulating its surroundings. We use Karel to solve the dilemma of either demanding too much or not enough from students during the first weeks of an introductory CS course, as interesting problems can be solved with limited input from lectures. Karel enables problem solving from day one of CS1, and encourages good software engineering practices such as top-down design from the beginning. We outline typical problems in the first weeks of CS1. We present a short overview of existing Karel implementations in various programming languages and our rationale for re-implementing Karel. We present our teaching philosophy and use of Karel in the classroom. We demonstrate how Karel is being used from a student perspective, along with a typical programming task. We discuss preliminary results of a survey and interviews with students from a first course in which Karel was used.

This contribution discusses problems of existing gamified learning platforms as a result of an analysis and reveals a research gap that is addressed using a domain-specific modeling (DSM) approach. Foundations of the DSM approach are described and our vision to use it for improving the creation, adaptability, and extensibility of platforms as well as the exchange of promising gamified learning concepts. Early results are presented and steps ahead are shown.

One of the biggest problems of educational games is the adequate integration of learning content into the game environment. Thereby the main challenge is to create a balance between gameplay and learning objectives and the correct transfer of learning principles into the game design. A useful approach is to identify game components and their specific learning constraints and the relationship between each other. By mapping learning elements to Game Design Pattern, we elaborate fundamentals for the concrete game design, focusing on card games. A shared description and an understanding of how the game design aligns with learning content facilitates the development of high quality educational games. A short illustration of our further developed implementation summarizes our experiences and shows an exemplary realization.

In this paper we investigate "bottom-up" gamification, i.e. providing users with the option to gamify an experience on their own. To this end, we review commonly used gamification elements in terms of their suitability for such an approach and present the results of an online questionnaire (N=75) complemented by semi-structured interviews with employees of a manufacturing company (N=8). In a twelve-day-long study (N=20) we investigated the usefulness of a task managing app implementing a "bottom-up" gamification concept. With these studies, we derived requirements "bottom-up" applications should fulfill. The study results reveal that people want to use such an approach and are open to the creation of their own gamified experience, thus suggesting that "bottom-up" can be an alternative to "top-down" gamification often used today.

In this paper we present a newly developed online learning platform which introduces gamification elements into software engineering education. Starting from assumptions based on cognitive load theory we present the design of an online gamification-based training system to be used in software engineering contexts. Students can voluntarily solve challenges for which they may earn credits. These small problems serve as assessments; the approach follows the assessment for learning paradigm in that assessments provide formative feedback to enhance the learning experience. The combination of formative assessment and gamification is new to software engineering education. We describe system design as well as the different types of challenges in detail. We also provide several examples for actual challenges used in an object-oriented programming introduction using Java.

Teaching software testing is a challenging task. Especially if you want to impart more in-depth and practical knowledge to the students. Therefore, most lectures still teach in a classic lecture format despite the fact that this way of instruction is in any case the optimal way of instruction for today's requirements anymore. In this paper we present our implementation of an active learning method to deepen the knowledge in academic software test education. We describe a card game for advanced learning that promotes students' collaboration and knowledge exchange in a playful and competitive manner. The design of the game is based on constructive and cooperative theories. A subsequent evaluation shows that the use of this card game for teaching software testing is a suitable method.

Within this thesis, we present our suggestions why playful learning in software engineering education is useful to mediate generic competences in academic teaching. Therefor we identified competences which are addressed by playful learning and mapped them to demanded generic competences in software engineering. Due to the well compliance, we analyzed current implementations of playful learning and their design regarding the mediation of required soft skills. Based on the lack of effective implementation, we close our paper with an exemplary design for playful learning.

In this paper the need for tool science, a discipline dedicated to the problem of developing, selecting, adapting and teaching about software tools for research is discussed. Starting from a general description of this field a short overview on the state-of-the-art is given. Core problems for tools research are discussed and several open research issues are identified like the need for case studies in research tool usage or making economic benefits of better usability and user experience for research tools evident. In addition, aspects of teaching concepts for tool developers and users outside the core disciplines of computer science and software engineering are presented.

In this paper we propose an augmented reality (AR) based assistance system for reliably teaching the assembly process of printed circuit boards (PCB) to workers by using a smart glass running a self-developed software. The system is operated freehand by looking at QR-Codes and highlights a component's retrieval location and installation point in the user's field of vision by using four markers. A study executed in a production line of an Electronics Manufacturing Services (EMS)-company resulted in an errorless performance of each individual participant who was equipped with the system. This paper describes the related work, concept and implementation of the software as well as the conducted study and its results. Finally a conclusion summarizes the success of the system and hints at future work.

This paper describes a case study of assessing student's coding behavior and skills in a realistic development setting. Students had to solve typical programming problems in the context of app development for the Android platform using the Eclipse IDE. Data was analyzed using IDE as well as browser interaction logs. In addition, screen recordings of the students' interaction with the IDE provide further insight. In this paper we present the first results of our ongoing work.

Current in-car navigation systems provide only a limited level of adaption to the driver and driving conditions. The driver's actual information need while interacting with the navigation interface is not taken into account. This paper aims at investigating the impact of situational features on the drivers' support need as well as proposing modes of adaptation for situation-aware navigation support. Therefore, three studies were conducted. It became evident that the driver's need for more or less navigation support depends on the complex interplay of different characteristics of the driver, the vehicle, and the driving environment. Based on our findings, a set of driving situation related adaption rules for a more user centered navigation support is proposed.

In the past years, driver analyzing has become a field of increasing interest. Within this topic, camera based as well as camera free systems are in the scope of researchers all over the world with the overall goal to detect, for example, critical driver states like drowsiness or distraction. Unfortunately, there are yet no comprehensive models for understanding the driver and his states in the automotive context. Therefore, we present a user model tailored to automotive needs. This model allows us to understand the driver in the automotive environment and to set up a general architecture from which we can decide on necessary input information for detecting a certain driver state.

This paper describes a software architecture approach which leads to a simplified solution for information retrieval in multiple backend systems. It is mainly based on the core ideas of service oriented and pattern oriented software architectures and supports the creation of simple structured, changeable retrieval systems with a unified user interface. It is mainly focused on the special needs of typical healthcare information system landscapes. The architectural approach has already been tested within a university hospital portal system which allows unified personalized access to patient dependent medical data to externals like related hospitals or resident doctors.