Fakultät Kommunikation und Umwelt
Refine
Document Type
- Master Thesis (37) (remove)
Language
- English (37)
Is part of the Bibliography
- no (37)
Keywords
- COVID-19 (2)
- Machine learning (2)
- NLP (2)
- PM2.5 (2)
- AI Hallucinations (1)
- AI integration (1)
- API-Driven Architecture (1)
- ASP.NET MVC (1)
- Agile (1)
- Air pollution (1)
Institute
Despite the remarkable achievements of large transformer-based pre-trained models like BERT, GPT in several Natural Language Processing (NLP) tasks including Sentiment Analysis (SA), challenges are still present for subdued source languages like Nepali. Nepali language is written in Devanagari script, has complex grammatical structure and diverse linguistic features. Due to the absence of balanced datasets, and computational resources for Nepali, achieving optimal result with the latest architecture remains challenging. For this reason, publicly available NLP modelsfor Nepali are very less, making research in this area difficult. This paper attempts to addressthis gap through the use of pre-trained transformer models specially tailored for Nepali from Hugging Face including BERT, DistilBERT, ALBERT, and DeBERTa for sentiment analysis in Nepali tweets on relatively balanced datasets. The models are trained on large Nepali datasets and optimized for NLP tasks involving Devanagari scripts. To evaluate the model’s performance, various tokenization strategies are investigated in order to capitalize on transformer-based embedding with the SoftMax function and confusion matrix. The outcomes of models are compared using the same datasets. The study’s results shows that DistilBERT achieved the highest accuracy rate of 88% in Nepali sentiment analysis tasks, followed by BERT and DeBERTa at 83% and 80%, respectively. However, ALBERT showed a low accuracy of 70%. The result of this approach shares valuable viewpoints for the field of sentiment analysis in diverse linguistic contexts.
The popularity of using transformer-based models like Bidirectional Encoder Representations from Transformers (BERT) for various Natural Language Processing (NLP) tasks is increasing rapidly. Unfortunately, the research is very limited for low-resource languages like Nepali. This study examines the utilisation of pre-trained BERT-based NERNepal for Named Entity Recognition (NER) tasks in Nepali text. The main goal is to investigate the efficiency of the NERNepal model, which has been pre-trained and fine-tuned on Nepali corpus data for NER. This research provides new insights by evaluating the NERNepal model on two distinct datasets. It addresses unique linguistic challenges specific to Nepali. The study also offers a detailed analysis of the model's strengths and weaknesses. By focusing on diverse datasets, this study shows how adaptable the model is and how its performance varies. These aspects have not been explored extensively before. The EverestNER dataset is one of the largest human-annotated datasets in Nepal so far, and the Nepali_NER dataset is also BIO-annotated for the NER task, which helped to compare the model’s prediction. The performance was better on the Nepali_NER dataset in comparison with another selected dataset. The EverestNER dataset contained many complex words for ORG connected with many tokens for a single entity with different contextual meanings and different annotations for the same word in different tokens as per context. Because of this, it created more confusion for the prediction, especially for the ORG entity. It had a similar issue with another dataset as well, but the label annotation was better in comparison. Furthermore, the research tries to clarify the difficulties and constraints related to utilising pre-trained BERT models for NER in low-resource languages such as Nepali. The study focuses on research areas on the efficacy of the model and its performance on two different datasets. Post-training with the EverestNER train dataset was attempted, but due to computational resource limitations, only a maximum of 3 epochs was possible, which did not improve the evaluation. It also has implications for enhancing language processing tools for Nepali. The results show the ability of a pre-trained BERT-based model to improve NER skills for Nepali text. However, further study is required to overcome the current obstacles to identifying complex words. The availability of high computational resources and the possibility of combining other NER approaches with a transformer-based model could increase the performance and robustness of the model.
Keywords: Named Entity Recognition, Natural Language Processing, BERT, low resources language
The recent development of generative artificial intelligence (AI) has inspired some game developers to try and integrate it into games. Anomalous behavior by generative AI or users seems undesirable and effort is being made to find ways on how to prevent anomalous responses by generative AI. However, these responses are a unique trait of generative AI and in our project, we explore strategies on how the negative effects of anomalous user and AI behavior can be mitigated, instead of trying to prevent them from appearing. We want to leverage their uniqueness as an intended game mechanic and see in what context these strategies can work. To achieve these goals, we researched already existing integrations of generative AI features in games and entertainment as well as possible connections to game mechanics and systems that are not related to AI. Based on this, we created 4 prototypes of conversational games with generative AI characters, which allowed the AI to create anomalous responses and did not limit the player freedom of what they may input. In our 36 participatory design sessions, we found that anomalous user and AI behavior can have a positive influence on the player's experience if a game's story is set up in a way that it can allow new and diverging story threads, that are created by the player and AI, to be integrated and if there are supplementary game mechanics, that allow the player to follow the intended story and complete the game.
During the past years, it is noticeable that the e-commerce industry has emerged drastically, offering accessibility to a variety of products to the customer where they can buy products from the comfort of their home. However, it is associated with a lot of new challenges for e-commerce businesses, particularly in understanding and managing the customer behavior patterns with the rising online product returns. Since the accessibility of online shopping has risen, assessing the critical factors related to product return and prediction has become a really challenging task for e-commerce vendors. This study also aims to segment the customers based on customer behavior prediction into two categories, i.e., high return risk and low return risk customers, and further develop strategies to reduce the online returns.
This study is rather structured into four integral parts, where each part provides the comprehensive analysis. The first aspect involves analyzing and identifying the customer behavior patterns leading to product returns. The second aspect predicts the online product return based on selected features and the third is to segment the customers based on online product return prediction and categorize them into high return risk, and low return risk customers. The final step involves development of strategies to reduce the product return based on the intense analysis conducted.
To achieve this meaningful research outcome, data analysis is conducted to understand the customer behavior patterns, and a Random Forest feature selector is used to identify the customer behavior patterns that lead to product return. Based on the identified features, classification models were applied to classify and predict whether the customer is going to return product or not. Furthermore, in these seven classification models such as Logistic Regression, Ada Boost, Decision Tree, Naive Bayes, XG Boost, K-Nearest Neighbors, and Random Forest were implemented and used to compare the performance of the classification models to find out the best performing model. Lastly, the segmentation of customers is carried out based on the online product prediction using Logistic regression with classification threshold method into high return risk and low return risk customer categories.
The results obtained help in understanding the customer behavior and reduce the online product return by developing the strategies. This study will eventually help the online businesses in reducing returns which will also enhance the customer satisfaction.
Data physicalizations are physical objects that encode data. The concept comes from data visualization, except for in a physicalization the data is presented as 3D tangible objects. Research shows that physical representation of data and being able to interact with it improve understanding and memorability of data, which supports the embodied cognition thesis, according to which cognition is supported by the body and the physical world and developmental psychology research, suggesting that manipulation of physical objects can promote understanding and learning. Alongside traditional physicalizations of statistical data, there also exist physicalizations of personal data, for example heartrate patterns, DNA data, one’s emotional states during the course of the week, etc. Research shows that such physicalizations, often embodied into everyday objects to promote familiarity and intuitive interaction, stimulate self-reflection as well as social interactions by telling others the stories behind the data. Some researchers thought about using data physicalizations as gifts and created example of such gifting artifacts. However, the design process of these artifacts was not based on the user research; nor were the physicalization gifts evaluated in any way. This is the gap that we are addressing in this master thesis, and the goal of our project was therefore to create and evaluate a personal data physicalization present using the human centered design approach. We started our work by conducting user research: performing interviews and defining the important characteristics of the future physicalisation gift. It was followed by a brainstorming session of possible design solutions and creation of the prototype. The prototype was then tested with users, addressing understandability of the data presented in the physicalisation and subjective perception of the gift. Our documentation of the findings and the research process provides a new step in the research of personal data physicalisations and lays out new topics for future investigations.
This thesis centers on enhancing the humanoid Pepper robot with Artificial Intelligence-driven information delivery capabilities in an educational setting using Large Language Model at student
service centers. In this work, Pepper was connected to GPT-Engine to act as a voice assistant robot.
With the help of GPT-Engine Pepper can answer questions about the University and its study programmes. The enhancement involves a client-server architecture, where the client-side application
enables students to interact with Pepper via speech, and the server-side, equipped with AI functionalities, processes these interactions and generates appropriate responses. This setup significantly
augments Pepper's capabilities, allowing for more accurate and efficient information delivery. An experiment was conducted with nine student participants of Hochschule-Rhein Waal, where they
asked questions about the University and study Programmes to inform themselves about the University and Study Programmes. During the interaction, they evaluated the robot's clarity of communication, accuracy of responses, and overall user experience. The findings indicate that while the AI enhancements generally improved Pepper's ability to communicate and provide information, areas such as voice recognition and response accuracy require further improvement. This research contributes to the field of human-robot interaction by illustrating a practical approach to augmenting existing service robots with AI capabilities. It underscores the importance of advanced AI integration in enhancing the functionality of service robots in educational environments. The study provides valuable insights into the broader application of AI-enhanced robots in the Information Guide role, highlighting the ongoing need for technological advancements in AI integration and humanrobot interaction.
Floods are the most common, devastating, and frequently occurring natural disaster nowadays.
Because of climate change, Europe is expected to see an even higher number of floods in the coming decade. As floods cannot be prevented, understanding the pattern and causes and being able to forecast could significantly reduce the losses during the flood. In July 2021, Germany faced a catastrophic flood, taking more than 180 people’s lives and causing around 40 billion euros of economic loss. To study the flooded region and the region at high risk, the Flood event 2021 has been visualized in this research. The Flood map has been created using a sentinel-1 image. In the visualization, floods in the Rhine River and Arh River can be seen, and the change in the pattern of flood from 12th July to 16th July has also been observed.
Different models exist that try to predict floods and warn as early as possible. Since Machine learning is popular nowadays in every sector, ML models have been implemented in this study to forecast the flood. The SARIMA, Random Forest, and LSTM have been implemented using historical data to predict the flood. The Random Forest model performed better than the remaining two models. The Isolation Forest model has also been implemented to classify the data into Flood and No Flood. This model performs well with the Flood classes but struggles in capturing the No Flood class.
API Driven Form Rendering
(2023)
This master’s thesis investigates the challenges and opportunities associated with dynamic form rendering in user interfaces, specifically in the context of Galeria Karstadt Kaufhof (GKK), a key player in the German retail sector. Collaborating with CODUCT Solutions GmbH, this research aims to enhance GKK’s operational efficiency and user experience in their digital transformation journey, focusing on Project Baldeney, a cloud-native platform that automates content management workflows.
The thesis begins by outlining the limitations of traditional, static methods of form rendering, highlighting the implications for development time and error frequency. It identifies the fragmentation in the existing API-driven architectures as a critical issue affecting system reliability and user experience. Against this backdrop, the work proposes a unified, API-driven framework to dynamically render forms for many user actions, such as data input and transactions.
The research takes a deep dive into existing systems—particularly Informatica’s Product Information Management (PIM) software and the new software in development Article Workbench to identify user experience issues, inefficiencies, and the complexities involved.
Through this examination, it offers a targeted solution to integrate various components of form management, like building, parsing, validation, and rendering, into a cohesive system.
The thesis aims to contribute to software development and user experience design substantially, by targeting improved development efficiency, reduced errors, and enhanced system reliability. By achieving these objectives, the research aspires to furnish GKK with
a robust solution for dynamic form rendering, thereby enhancing both user satisfaction and system performance in the fast-evolving digital landscape.
The literature review at hand addresses the following research question: “Does artificial intelligence and its future development potentially pose a significant threat to individuals’ well-being and humanity’s ethical values?”. For providing answers to the topic at hand, three hypotheses are investigated in detail out of which H0: “Artificial intelligence does pose a significant potential threat to individuals well-being and humanity’s ethical values.” is accepted in accordance with evaluation of findings gathered from 121 sources overall. Threats of artificial intelligence were identified in overall five fields of application, namely medical science, transport industry, potential militarization, surveillance as well as in interaction of artificial intelligence with consumers, for which the natural language processing model ChatGPT was investigated. Regarding medical sciences, threats include potential dehumanization effects through care robots equipped with artificial intelligence. In the transport industry, multiple issues may arise with autonomously operating vehicles while in potential military application of artificial intelligence, research is conducted in multiple countries. Regarding surveillance, artificial intelligence severely facilitates the implementation of social credit systems, as for instance seen in China. Additionally, human nature is embedded in the context of continuous development of artificial intelligence with aim of establishing a first approximation to guidelines of potentially needed limitations of further development of artificial intelligence.
It is concluded that, for implementing specific limitations on development of artificial intelligence, more research is needed for identifying the most suitable approach of limiting further development of artificial intelligence.
The modern work environment is subject to constant change with a notable shift from individual to team-based work structures. This transition necessitates an improved un-derstanding of how managerial actions influence the dynamics within these teams, to optimally align them for success. Despite the substantial research about leader effec-tiveness and team effectiveness based on organizational performance indicators, it is rather infrequent that the direct influence of leader behavior on teams is explored. This thesis aims to address this gap by answering how leader behaviors impact team pro-cesses in the context of German medium to large-scale IT-enterprises. Using a quanti-tative cross-sectional methodology, individuals working in team-based structures under direct supervision were surveyed. Analyzing data from 94 respondents via hierarchical linear regression analysis, several significant relationships were identified. In general, leader behavior has a strong positive impact on team processes. Furthermore, task-oriented leaders have the greatest positive influence on action processes, while relation-oriented leaders have the greatest positive influence on interpersonal processes. Change-oriented leaders also significantly influence action processes and interpersonal processes, albeit less strongly than their counterparts. These results provide implications for team management practices, emphasizing the importance of leader behavior for aligning team processes in the direction of success.
Groundwater is an essential resource that is used for a wide range of purposes all over the world, and retaining this important natural asset demands that it be managed and utilized sustainably. Despite the fact that groundwater is frequently abundant and widely utilized, it is difficult and costly to precisely estimate water levels when compared to surface water systems. Groundwater managers require improved digital tools to understand the system state of the aquifers they seek to interact with sustainably (e.g., accessing real-time updates on the state of the water table to check whether
pumping operation is achieving or violating a desired drawdown target.) Here we propose a workflow for a pilot region that (a) temporally interpolates irregularly, manually-measured water levels in a real observation network using data from neighboring sensor-equipped observation, using a multiple linear regression approach to
reduce complexity. The resulting interpolated time series are then used in conjunction with the sensor data to come up with (b) spatial interpolations of the groundwater field over time using inversed distance weighting (IDW) at any given point in time. From
that, spatial estimates of the groundwater field can be transformed into maps of deviation from drawdown targets. We visualize these targets and their temporal evolution for groundwater managers in real-time on a web application. The web application displays these maps together with operational data (e.g., extraction rates)
and meteorological data (recent and forecasted precipitation rates). It serves as a foundational tool for groundwater management when making control decisions to save energy in climate change mitigation and to reduce costs.
Brand communication through social media has grown significantly in recent years specifically during the covid-19 pandemic, becoming one of the most popular and effective forms of online communication for businesses and consumers. Due to the tremendous increase in social media brand communication during the COVID pandemic, businesses need to study the views or the response of consumers to it and understand how to fully leverage it.
With the objective to understand the consumer's views toward social media brand communication during the pandemic, influencing factors of social media brand communication were first examined, and then consumers' responses towards those factors were studied. Two research questions were developed to execute the study.
The thesis first relies on a literature review to assess the influencing factors of social media brand communication during Covid-19. Then the thesis, in the empirical part, applies the qualitative research method by conducting semi-structured interviews with 10 participants grouped by aged 25-34 and users of both Facebook and Instagram. The interview was conducted via ZOOM, audio recorded, transcribed, and analyzed using MAXQDA.
The result showed that the influencing factors of social media brand communication were informativeness, entertainment, credibility, interactivity, and electronic word of mouth. In general situation, the brand’s product information and value-related information were the central focus of the consumers but during the covid pandemic covid related information, CSR, health-related messages in entertaining form, source expertise, high level of brand interaction and word of mouth from the trusted sources were the influencing factors of social media brand communication. Consumers responded favorably to those brands that were able to incorporate all these factors in social media brand communication.
The study is limited to Facebook and Instagram users in Germany who are between the ages of 25 and 34.
Personality Traits and Work Engagement among Human Services Workers: The Role of Coworker Support
(2023)
It has been increasingly important to study the well-being of workers in the human services industry as the society benefits from their care and effort abundantly. Most research on human services workers has been focused on the topic related to mental “un-well being” such as exhaustion and burnout, instead mental “well-being”. However, lately a more positive approach of work engagement was introduced as a form of positive psychology. Since previous studies of work engagement have focused mainly on its environmental factors, this study attempted to explore the construct from another perspective which is the personality differences (extraversion and neuroticism). The main objective is to answer the question of why some individuals are more engaged than others while working in the same conditions. Moreover, the moderating effect of perceived coworker support as a form of social support is also examined in the model. A total of 246 nurses and caregivers working in Germany participated in an online survey. Findings show that extraversion and neuroticism are significant predictors of work engagement which indicates that extraverted workers are more engaged compared to the ones with high level of neuroticism. Coworker emotional support was found to buffer the negative effect of neuroticism on work engagement. However, there is no evidence of a moderating effect from coworker instrumental support on the relationship between personality traits and work engagement. The result of this study can be useful particularly for practitioners who may benefit from designing training programs that foster helping behavior among workers hence improving their work engagement.
Companies often utilize error taxonomies to handle software problems. Agile and scrum-based software engineering procedures leverage software testing and quality management to find, repair, and prevent functional faults. Taxonomies help programmers see patterns and solve comparable challenges. Usability engineering-based taxonomies emphasize the use of HCD approaches to discover and classify usability concerns across the software development life cycle. Software and usability engineering have pros and cons. This thesis presents a hybrid software challenge taxonomy to bridge these two areas. The research project aims to build a new hybrid error classification system for identifying and correcting software errors at their source, leading to more user-centric software. Software engineering taxonomies and design issues are investigated. The weaknesses of any research taxonomy are then evaluated against one that fits some of this thesis' needs. User interviews and contextual inquiry are used to uncover software faults. The results of the study showed that poor usability, a mediocre user interface, workflow issues, and poor user experience caused most non-technical problems. Non-functional faults may have created many technical challenges. Six out of ten company professionals who participated in a questionnaire and interview mistook "usability" for "user experience." Usability and user experience were valued by two out of ten specialists, but they were misunderstood, expensive, and time-consuming premium approaches, and thus not included in the software development lifecycle. Ignoring non-functional issues may have caused numerous software failures. Scrum and agile used alone may also be to blame for many of the software issues uncovered in this thesis project.
This thesis explores how UXD and SD can work together more seamlessly in practice.
UXD and SD have a lot of similarities. This has led to confusion around the borders and overlaps of both fields. Therefore researchers called for more studies to clarify their relationship. Most recent studies suggested diverting from polishing single touchpoints; some studies and focussing on end-to-end journey and the customer.
The thesis examines the underlying activities of existing theoretical approaches and the challenges of combining UXD and SD with the help of thematic analysis. A set of proposed strategies were developed based on a survey on current practices and challenges with practitioners and thematic analysis. As a result, the thesis presents four strategies that design teams can incorporate into their processes.
The idea is to promote the creation of better quality products and services by harnessing the benefits of both fields. The key point was that these concepts are still broad and should be simplified into leaner modes that can be applied in daily activities.
The end of 2019 has brought upon the world a new virus called Covid-19 that has quickly evolved into a pandemic at the start of 2020. To control the spread of the virus, countries around the world were forced to implement various strict measures that includes travel bans, lockdowns, and limited social contacts. As per pervious pandemics, this crisis had severe consequences on economy, consumer behavior, purchasing decisions and people’s mental and physical wellbeing. Consumers were forced to adapt and find alternatives that would fit the “new norm”. This added to the psychological distress people were experi-encing and in turn increased their feeling of risk and uncertainty. Thus, this research con-ducts a semi-structured exploratory in-depth interview (N=16) to profoundly investigate the effect of risk and uncertainty on consumers’ purchasing decisions in Germany. Find-ings of this study suggests that consumers with high uncertainty avoidance and high per-ceived risk are more likely to have a changed behavior. Their purchasing decisions are more oriented toward necessities and self-development products. On the other hand, con-sumers with low uncertainity avoidance and risk perception have experiences less change in terms of consumer behavior and purchasing decisions. The results of this research would help companies understand consumers’ psychological needs which is often ne-glected in literature to provide the needed products and services. It would also benefit the academic and professional world by knowing what to expect during crisis and if changed behaviors will continue after the pandemic is over.
An exploratory study of how to effectively communicate the level of hotness/spiciness of food items
(2022)
Preferences for spiciness in food varies greatly from person to person. In addition, the notion of spiciness itself varies greatly from culture to culture and from region to region making it hard to effectively represent and communicate the level of spiciness of food items in restaurant menus and food packaging. In this study we explore the use of virtual taste sensation as a means of effectively communicating the desired or available level of spiciness between the consumer and the provider. Based on prior research on electrical and thermal stimulation of taste sensations, we have developed a device capable of simulating the sensation of hotness on the human tongue through the use of a low voltage electrical current delivered via a set of electrodes and control panel. The development process of the device consisted of the following two steps: 1. We conducted an observational study in which participants were asked to rate their spiciness preference using a commonly used hot sauce which was mixed with either water or rice. 2. Participants were asked to match their spiciness preference using the device at different voltage levels. Although the device only generates the stinging sensation and not the full flavour sensation associated with a spice such as a chili pepper it proved effective in assessing and communicating the spiciness preference of the study participants. Thus, we argue the device holds great promise for developing applications focused on communicating spiciness levels, e.g. the spiciness of the contents of a food package, the spiciness of items in a restaurant menu, or the spiciness of various chili peppers in a market.
DevSecOps has become a new paradigm in the technology world in a very short time. Despite the fact that there is a lot of talk about it, but in practice it is much more complicated. This paper shows one way to automate security testing on an established client-server application.
This research aims to find out if it is possible to improve quality standards by automating security testing and reducing human effort. One of the goals is to find reliable sources on information security management and practices known today. The next goal is to determine which components of the application to perform security analysis on.
We took the Open Web Application Security Project (OWASP) as a reliable source and began to examine the Top 10 web application security risks. The research needed to understand how this could be further applied to our client-server application.
As a result, we ran a manual experiment on the application following the OWASP Top 10 guidelines. After achieving some results in the manual experiment, we already had an idea of what to test and tools to automate.
We managed to automate 4 out of 10 categories, which took us about 2.5 months. On average, automated testing takes 38 minutes across 4 categories.
In conclusion, it should be noted that automating security testing costs a lot of effort and resources. Based on the example of our work, we can say for sure that it is necessary to properly prioritize resources and smoothly move to automation. The approach to cyber security analysis and automation will be different for each project. Nevertheless, information security in the vastness of the Internet was, is and will be, regardless of how technology evolves.
This thesis investigates the user involvement practices in the product design process of FinTech UX practitioners mainly located in Nigeria. The study aims to identify the design activities used to involve users, the methods the practitioners used, the challenges they faced, and how they measured the impact of these activities. The study used an online survey research method to collect quantitative and qualitative data from 62 respondents. Results from the survey show that user involvement is mainly done during testing. However, it is less prevalent during the software design activity. Some of the significant challenges reported are user recruitment, user participation, user requests, budget, and time constraints. A mix of direct and indirect user involvement methods were also reported, such as surveys, observational techniques, user interviews, data analytic tools, user communities, and social media platforms. To measure the impact of user involvement, the practitioners mainly depended on analytical data, user feedback, and reviews.
For most of the practitioners, user involvement is not well established in their design process possibly due to the project-related, user-related, or organizational-related constraints reported in this study.
News is one type of information that has the potential to influence a large number of people. People have received news since the beginning of time through various birds, short letters, and other means. When newspapers were invented, information was available everywhere on paper. The news is no longer limited to paper-based platforms, thanks to the digitalization of online platforms. The online news platform is now available to read the news in a matter of seconds. As a result, news can easily connect people, and it is being used to spread fake news. Fake news is spread to gain attention for the wrong reasons. Different languages are used in our world to express our thoughts and feelings. There are specific materials for each language. The English language is the most studied topic when it comes to identifying fake news. Data and research resources are few in other languages, hence there is little research done. Of these, Bengali is one of the most widely spoken. Our ultimate goal is to create a tree that contains elements of both English and Bengali. Research on fake news and web scraping was used to get the language news data. Multilingual transformer models m-BERT and xlm-ROBERTa with long text or tokens are used to detect fake news (512 tokens or any token size). The two models were compared using two different datasets (with stop words and the other without) using three different fine-tuning freeze approaches (Freeze, No Freeze, and Freeze Embed). The results show that the dataset with stop words had a somewhat better performance than the dataset omitting stop words. The xlm-RoBERTa model outperforms the m-BERT model in terms of F1-score and accuracy.
This thesis addresses the question of a suitable way to communicate user requirements to an interdisciplinary team. UX professionals are responsible for communicating user needs and UX requirements in a clear and understandable way so that other team members such as software developers can work in a user-centric way. The thesis examines known forms of requirements specification and identifies possible challenges in communicating user needs with the help of a systematic literature review. Based on design patterns, possible solutions for these problems are developed. In an interview study, these solution proposals are compared with real scenarios and adapted. As a result, the thesis presents nine guidelines that UX professionals can use to communicate user requirements. The idea is to promote a good exchange with the other team members and suitable, short forms of specification. A key point was that these should always convey a reason or the user’s motivation for a requirement to provide meaningful context information to those who have never been in touch with the user.
Various companies have been collecting people's digital footprint by offering free digital services to them. People are aware of it vaguely, it is difficult for them to see what actually the companies have or its value. According to the General Data Protection Regulation, finally, individuals can request their collected data from the companies. Nevertheless, the data has been offered as machine-readable files mostly that individuals cannot understand. Thus, a data visualisation application for individuals was designed to improve these problems in this project. The application was built as low fidelity and high fidelity prototypes and tested with 10 participants through remote usability tests to explore its possibility. In the tests, the target users showed their positive interest in the concept and visual design. Consequently, it clearly showed its potential to improve the situation. However, some participants seemed to worry about sharing their data for donating, or unsatisfied with the basic level of data visualisation. To cope with these issues and to have a better effect, some improvement of interface design, additional introduction screens and advanced data visualisation was suggested as a further improvement for the future.
Participatory Virtual Reality. Immersion and presence in digital platforms for serious topics.
(2021)
Energy is one of the most important components of our lives. Nevertheless, it is still considered an abstract concept for a great majority of the population. Energy is also responsible for most of the environmental problems we are facing nowadays, therefore, we require more than ever, and by any necessary means, to bring this concept closer to our daily life, not only from a dimension of knowledge, also from those related to our behavior and attitude towards it. This study aims to explore Virtual Reality social platforms as a promising means of communication to address environmental issues. More specifically, to understand their potential to generate participation cultures around serious issues, understanding participation culture as the set of activities, through which people collectively shape a space for expression and learning. For this purpose, a research was conducted analyzing three Virtual Reality social apps (VRChat, Rec Room, and AltSpaceVR) to identify —under the perspective of Jenkins’ Convergence Culture— the existing activities and behaviors shaping participatory cultures inside of the platforms. A second research was conducted based on the observation of the main components of the platforms to use them as a reference for the conceptual development of a VR app to present the strategic plan of the European Commission in a participatory environment.
Natural gas is an indispensable source of energy for industrial and domestic purposes, hence distribution and trade demands competition. In gas distribution, optimizing is a crucial task involving control decision on the network elements with respect to trader’s demand at the boundary nodes. The application of artificial intelligence is becoming popular in diverse domains, making its role in gas transport networks more and more interesting. The opposite interest of the players in the network is the reason behind the thought of establishing a game analogy here. The agents as players act on the basis of the opponent’s action. Since both agents perform actions from different space, a neural network model is designed for the dispatcher agent to learn to make decisions for active elements based on its experience from the sample gas transport network. Behavior of the network and its elements is depicted using an optimized gas network simulator. The representation of the gas transport network and its state is digitally transformed so that it is able to extract the features by an artificial neural network representing the environment and its rules. With the help of the trained neural network, the dispatcher agent decides the control decisions with less accumulated penalties. The results and findings of the proposed method are subjected to a comparison with the interval halving method. The results of the study will open the door for further research possibilities.
Air pollution has been one of the main concerns for Nepal for a very long and PM2.5 is one of the most neglected air pollutants in Nepal. Nepal is among the countries most affected by both indoor and outdoor PM2.5 pollution in the world. This thesis is based on the recent lockdown that happened in Nepal during the pandemic of COVID-19 in 2020. This study investigates the impact of lockdown on PM2.5 emission based on the assessment of emission inventories. The annual data for PM2.5 emission was extracted from EDGAR (Emissions Database for Global Atmospheric Research) and then the extrapolation was done to calculate the assumed PM2.5 emission for 2019 and 2020 using linear regression. The assumed nationwide monthly emission ercentage was used to calculate the PM2.5 emission during lockdown months. The PM2.5 sources affected during the lockdown were identified and their assumed percentage contribution (6.18%) was used to calculate the emission during lockdown month. The ambient air PM2.5 concentration data was extracted from the official websites of the government of Nepal. The lockdown months were compared with the preceding year's same month and the percentage change was calculated. The study suggested that in 2020 the PM2.5 emission rate for April was reduced by 4.68% and for May it was reduced by 4.85% whereas during an assumed 'no lockdown‘ year, all the months would have had an increase in PM2.5 emission by ~1.6%. The concentration of PM2.5 for March was educed by 14.65%, similarly, for April it was reduced by 13.96% and for May it was reduced by 50.92%. The comparison of the percentage change of both the PM2.5 emission and concentration shows that the decrease in concentration was much higher compared to the reduction in emission due to the lockdown. The monthly assessment of PM2.5 during the lockdown suggested that the percentage change in emission of PM2.5 was minor compared to the change in the concentration of PM2.5 which might be due to different reasons including meteorology, underestimation of the contribution of some emission sectors by the emission inventories and the fact that the emission inventories are set up for the whole country whereas the PM2.5 concentration data are collected with monitoring stations located in cities close to major PM2.5 sources.
Language models are widely used as a representation of written language in various machine learning tasks, with the most commonly used model being Bidirectional Encoder Representations from Transformers (BERT). It was shown that the prediction quality strongly benefits from language model pre-training on domain-specific data. The publicly available models, though are always trained on Wikipedia, news or legal data, thereby missing the domain specific knowledge about medical terms. In this thesis, we will train a BERT language model on medical data and compare performance with domain-unspecific language models. The dataset used for this purpose is the Non-technical Summaries - International Statistical Classification of Diseases (NTS-ICD) task of classification of animal experiment descriptions into International Statistical Classification of Diseases (ICD) categories.
The COVID-19 pandemic has drastically changed the lives of every human being in most of the countries in the world. Lockdown has been implemented in many countries to prevent the virus from spreading. Lockdown has prohibited people to go outside of their houses, closed industries and factories, and restricted air transportation and the free movement of vehicles, except emergency vehicles and flights. On the optimistic side, improvements in ambient air quality due to the lockdown have been reported across the world. This study evaluated ambient air quality by analyzing the PM2.5 concentrations monitored in 7 major cities of Nepal during March, April, and May of the years 2019 and 2020. Furthermore, meteorological data were also evaluated to investigate how meteorological conditions have affected Nepal’s ambient air quality during the investigation period. The monthly mean PM2.5 concentrations during March, April, and May 2020 were found to be lower in nearly all of the 7 cities of Nepal as compared to March, April, and May 2019. To some extent, a slightly higher amount of precipitation in March-May 2020 than during March-May 2019 supressed the ambient PM2.5 concentrations in 2020. But, since the highest decrease in PM2.5 concentration was found in Nepalgunj (72.78%) followed by Kathmandu (46.2%) during May 2020 as compared to May 2019 and thus, the highest differences between 2019 and 2020 were found for the cities with the highest contribution of transport emissions to PM2.5 in ambient air, the COVID-19 lockdown might have contributed to the lower PM2.5 concentrations in 2020. On the other hand, it can be assumed that one of the major primary sources of PM2.5, biomass burning (both the residential use and natural forest fires), was not affected by the lockdown, but biomass burning might be the reason for the increment of PM2.5 concentrations found at some sites, in particular in the Kathmandu Valley, and for the only subtle decrease at other sites during the lockdown. In essense, though it is an unfortunate situation, the lockdown provided a glimpse of the situation how effective improvements in the environmental conditions could be if we were to introduce large scale mitigation measures swiftly.
Keywords: COVID-19; Lockdown; Air Pollution; Nepal; PM2.5 Concentration; Meteorological Data
Purpose — The purpose of this thesis is to evaluate the perceived effectiveness of the Design Sprint process according to practitioners in the field, and investigate optimal contexts of use as well as the limitations of the process.
Design/Methodology — A qualitative study based on interviews conducted with eleven practitioners from startups and large enterprises. Data from the interviews were analyzed, interpreted, and related to the existing literature in order to answer the research questions.
Findings — The hypotheses regarding the effectiveness of Design Sprint and its role in enabling efficient and effective decision-making were validated by the eleven practitioners who took part in the research. Additionally, patterns regarding optimal contexts of use of Design Sprint were identified, namely kickstarting product development initiatives and creating innovative solutions, as well as limitations of Design Sprint such as its limited use for exploring the problem space of challenge.
Limitations — The research is limited by the small number of eleven participants and the lack of quantitative data.
Practical Implications — The study provides useful information to researchers as well as practitioners enabling them to understand the effectiveness of Design Sprint in real-world contexts. The data from this research can be used to further investigate the topic of Design Sprint in the literature while providing practitioners with insights to help them decide in which contexts it would be best to deploy Design Sprint and gain internal support for using Design Sprint.
Originality/Value — This is the first qualitative research that investigates the effectiveness of Design Sprint and academically validates its claims of value.
Keywords — Design Sprint, product development, innovation, process
A steady decline in voter turnout has been recorded in many democratic countries, with young people (i.e Gen Z and Millennials) identified as a significant portion of those who don’t vote. Research has shown that one of the reasons young people do not vote is a lack of information about the electoral process, candidates and policies. Studies have also shown that preferred information dissemination medium is aligned across age groups, with older people trusting old media such as newspapers and young people leaning towards new media such as social media. With the recent advancement in machine learning, natural language interface and popularity of conversational agents such as chatbots, a new subset of new media has emerged. This study aims to investigate if chatbots could be an efficient medium for young people to access electoral information. It does this by comparing it with a website, using the EU electoral page as a case study.
Based on the content of the EU electoral webpage, a chatbot was designed, trained and deployed on Telegram. Participants were randomly divided into two experimental groups (chatbot and website) and asked to find a list of crucial information, interact with the medium and then respond to the System usability scale (SUS), and a perception questionnaire. Observed data and interview notes were also taken. Analysis of the experimental data, observed data and surveys demonstrated that chatbot as a tool is inefficient as an electoral information tool because of the limitation of chatbot when out of scope questions are asked and limitation in following conversation contexts which is crucial to conversations. However, it is rated as above average on usability and joy of use, but raises questions on trust and data protection. It is recommended that the study be performed again when there are more advances to natural language processing to see if its efficiency as an electoral information medium will improve. A longitudinal study is also recommended to see if there’s a direct impact of chatbots as a tool, on young people’s motivation to vote.
This research focuses on measuring learning curves from different people while they cook with the aid from various media channels. For such purpose profile inquiries, face-to-face interviews, self-recorded assignments, and classification tests will be used. The analysis of the results intends to derive into good practices of e-learning media design.
This documentation within the context of my master thesis is
about my virtual reality installation called “Synagogue of Kleve”.
The installation deals with the reconstruction of tangible cultural
heritage in virtual reality. The installation is conceptualized to be
shown in the cultural heritage institutions, including museums
and historical sites in which the visitors explore the synagogue
of Kleve, Germany.
The result is an installation that shows the synagogue in old
times, and that strives to get people interested in the history of
the synagogue with the help of virtual reality that features an
interactive storytelling experience.
Stored Procedures perform a significant role in the implementation of complex business logic in the database management system. Web-based application systems that handle a large amount of data with complex calculation and transformation rules, they prove to improve the system efficiency, maintainability, data integrity and reliability to a large extent.
This research will attempt to enhance a FoxPro based timesheet management system by implementing a web-driven Relational Database Management System with SQL based stored procedure and user-defined functions for faster and efficient data processing. The thesis explores the issues and the processing stages of the current timesheet management process and develops user-friendly responsive intranet online web-application portal with minimal processing steps. That accomplished by writing the processing logic using stored procedures in association with the user-defined table-valued and scalar-valued function in SQL Server and integrating them with Entity Framework three-tier web application architecture in ASP.NET MVC. The final timesheet booking data generated by the stored procedure is modified to a specific format using SpreadsheetLight, an open source spreadsheet library for the .NET framework for uploading the output booking file on SAP Business ByDesign web portal. Outcomes show that using stored procedures in a comprehensive business application can enhance the application performance by lessening the processing time and processing stages.
Purpose: The present study aims to investigate the acute effects of mental practice on novices’ basketball free throw performance and their motor self-efficacy. Moreover, it is examined whether internal visual imagery combined with kinesthetic elements enhances performance more than external visual imagery.
Methods: In order to test the research questions, a sample of 61 participants who are currently not participating in any type of basketball program were assigned to one of the following three conditions with the help of the matched sample method: The internal visual imagery group with kinesthetic elements, the external visual imagery group, and the control group. The experimental interventions with a duration of 15 minutes (mental imagery internal or external or mathematic exercises) followed by physical practice were repeated three times in a row. Consequently, the basketball free throw enhancement was measured via the shot accuracy. The results were compared by using an ANOVA with repeated measurements. Changes in the participants’ motor self-efficacy were assessed via a pre- and a post-test of the MOSI (Wilhelm and Büsch, 2006).
Results: Results showed no significant differences between the groups, neither in the performance increase nor in the self-efficacy.
Conclusion: The current study can serve as a reference point to avoid overestimating the effectiveness of mental practice. Findings are discussed in terms of possible explanations and hints for future research are given.
The urgent need for plasma and its components is increasing rapidly all over the world. This is due to demographic changes, which bring about an imbalance between eligible donors and possible future recipients. Hence, these changes pose a great challenge in meeting the demand for plasma and plasma-derived products in the future. As plasma cannot be synthesized, recipients depend on voluntary plasma donors. In order to ensure a sufficient supply of plasma considering the ongoing process of aging population, it is crucial to understand why people engage in donation behavior and which factors influence their decision to maintain the donor career.
The current study investigates the determinants of plasma donation intention using an extended version of the theory of planned behavior. As the study is conducted in cooperation with Octapharma Plasma GmbH, a subsidiary of one of the largest worldwide operating providers for plasma preparations, the analysis is based on the company’s donors only. By using an online questionnaire, data of N = 1153 plasma donors is surveyed in order to examine the hypotheses. Multiple hierarchical regressions reveal significant predictors for plasma donation intention for all donors combined and also for both first-time and repeat plasma donors. Across all donor stages, self-efficacy turns out to be the predominant predictor for future donation intention. Moreover, the satisfaction with the last donation influences all groups of plasma donors in their decision of career maintenance. Therefore, organizations should pay attention to these components, developing strategies to enhance both donors’ self-efficacy and donation satisfaction.
The German logistics company Schenker AG categorizes its customers into so-called vertical markets. For instance; category “Automotive” is assigned to car manufacturer BMW AG. The classification allows the company to evaluate its revenue and profits on different customer segments which, in turn, has an impact on Schenker’s strategic planning.
Until now the assignment is carried out manually which means that someone from sales department should perform some research on a customer’s public profile whenever a new customer is registered in the database. With the rapid growth of global trade in recent years and Schenker’s expansion to the Asian market with thousands of new customers the manual approach is no longer sustainable in a global market.
This thesis provides an alternative solution based on scraping customer data available in the web and classification of the extracted content. We deal with three significant difficulties: find web data related to a given customer name (we do have a company name but no homepage URL in Schenker database), extract a predictive portion of the data without introducing too much noise and, finally, set up a classification algorithm. Most importantly, the whole process needs to be implemented automatically.
For the classification task, we have identified two tree-based classification algorithms random forest and extreme gradient boosting (xgboost). Random forest performs better by using package ranger with an overall 52% accuracy and 87.8% multiclass area under the curve. On the other hand, xgboost takes less time to compute, but the accuracy is poor as compared to random forest.
Usability is an important quality aspect of software products. Today, companies need to be capable to develop software with good usability to stay competitive. This can be achieved by practicing human-centred design (HCD) during the development process. However, especially small and medium-sized software enterprises (SMEs) still practice HCD on a low level, if at all.
This thesis explores whether and how the approach of process capability assessment can support SMEs with a systematic improvement of HCD. It is shown that a process capability assessment for usability at SMEs can only be successful, when the first steps of initiating HCD are al-ready done. Then, a process assessment approach seems to be useful to systematically improve the usability capability at SMEs.
Based on this theoretical background, a usability capability self-assessment tool tailored for SMEs is developed. It has a simplified assessment procedure and uses the DIN EN ISO 9241-220 as reference model. The self-assessment tool was applied at two German software SMEs to test whether the approach is useful for SMEs and whether the reference model is suitable.
The results indicate that the usability self-assessment tool is a promising approach for future SMEs as several interesting applications were identified. The chosen reference model seems to be an appropriate choice for a self-assessment; only the complex language used in the standard impedes the assessment. All in all, the thesis lays the foundation for a systematic approach to support SMEs with usability.
This master thesis presents a machine learning approach using the distributed data- processing framework Apache Spark and the programming language R. The aim is to predict the time series of the Appl Percentage (ApplPerc) from the workload manager of the z/OS mainframe system using SMF 72.3 and SMF 70.1 records as input. System Management Facility (SMF) data are binary log files that are used to collect system performance data and information about system behaviour. The SMFs show how Apache Spark can be used for the pre-processing. This includes data collection, data extraction, and data selection. The machine learning algorithms are implemented in R. The input SMF datasets will be split into training and test data, and thereby applied to different machine learning and deep learning models such as random forest regression, recurrent neural network, and k-nearest neighbor regression to predict the ApplPerc. The validation of the prediction models will be proved by using cross-validation techniques to evaluate the best applied parameters for each model and therefore to locate the model with the best performance. The process of data analysis is followed by a data-mining methodology called Cross-Industry Standard Process for Data Mining (CRISP-DM), which outlines the steps involved in performing the analysis.