The goal of the bachelor thesis is to create a comparable prediction quality by using a generalized methodology for data preprocessing and the same prediction model for industrial and media applications.
Supervisor: Jun.-Prof. Dr. Matthias Hirth
External advisors: Dr. Alexander Graf and Hr. Fabian Hainzl |PFI | P (ZF)
Within the scope of the bachelor thesis, recommendations for action are to be developed for greater acceptance and cross-generational cooperation in the media industry. The work will be carried out using a case study and in cooperation with the production and broadcasting operations of ZDF in Mainz. The field of tension of the topic is, on the one hand, the age structure of the employees and, on the other hand, progressive technological innovations and challenges that shape the everyday work of the job profiles of the production and broadcasting operations. The analysis focuses in particular on the effects of new technologies on productivity and effectiveness in everyday working life, the willingness to use new technologies in everyday working life, the differences in dealing with new technologies on the basis of demographic characteristics and the extent to which technological challenges are mastered across generations.
Supervisor: Dr.-Ing. Mathias Bauer
The Thinking-aloud protocol (TAP) is still one of the most relevant and widely used usability evalaution methods. TAPs are usually conducted in usability labs, which allows for monitoring tests with specialized equipment. However, fully equipped usability labs are expensive, the tests take time and may only be accessible to local testers. In this context, crowdsourcing could provide the opportunity to perform TAPs in a time and cost effective way while reaching a global audience. This approach implies the absence of test facilitators, which makes it difficult to monitor participants and the test environment in real time, i.e. making it challenging to think aloud. The goal of this thesis is to design and evaluate a crowdsourcing based framework for the implementation of TAPs.
In this thesis, machine learning techniques were used to understand the users' behaviour and predict their engagement on the exemplary social media platform Boombazoo. The users' behaviour was analysed using the k-means clustering algorithm. Three different clusters emerged of the sample users: Passive observers, Potential creators, and Incognito creators, each with different demographics, social circles, and frequency of using the Boombazoo. Further, different classification algorithms were used to predict the users' engagement. Most of the classifiers had similar results on predicting the user engagement trend with an accuracy of around 74 %.
The customer service of a health insurance office deals with a large variety of tasks. In addition to complex consulting tasks, these duties may also include simple things such as printing forms for customers. This thesis describes how a self-service terminal has to be designed and whether it meets with acceptance within the target group. Furthermore, it is examined what influence the form of presentation has on it. In the course of the thesis two design variants are evaluated with real customers in a field study over a period of four weeks and the acceptance of the system is evaluated. The results are promising and show that even elderly customers without prior computer experience would use such a terminal. Furthermore, the results show that users feel comfortable even with a minimalist design, which enables the development of such a product in a simple and uncomplicated way.
Supervisor: Jun.-Prof. Dr. Matthias Hirth
Crowdsourcing enables an individual or an organization to propose the assumption of tasks to a heterogeneous group of people via flexible open calls. The acceptance of tasks leads to mutual benefit because the crowdworkers receive some kind of reward, e.g. economic or social recognition, and the crowdsourcer benefits from what the crowdworker has done. Typical crowdsourcing tasks include writing articles, tagging content, conducting user surveys and completing and correcting data. The quality of the results of a task depends mainly on the crowdworkers, so their motivation is a relevant factor to consider when designing a task. In this context, gamification, i.e., the use of game design elements in non-game contexts, could be used to promote the motivation of the crowdworkers. The goal of this thesis is to design and evaluate a gamification strategy in the context of crowdsourced image annotation.
The main purpose of this media project is to optimize the training of crowdworkers for the annotation of histological images. To aim this, a breast cancer image annotation tool will be developed. Then, error patterns during the annotation process will be identified using the Thinking Aloud method. The identified patterns will be used to optimize the training process of crowdworkers and improve the annotation tool iteratively. We expect that the optimization will lead to improve the quality of the annotations.
The CrowdTA framework is a Usability Testing approach that attempts to combine the benefits of the Think Aloud Method and Crowdsourcing . However, the uncontrolled settings of Crowdsourcing lead to challenges in the collection of high-quality verbalization from crowdworkers during the unsupervised online testing sessions. Typical scenarios would be insufficient audio level, interference noise from broken equipment, background noise, or the wrong language usage. The current project aims to develop tools and systems to detect low-quality Indicators in crowdsourced verbalizations collected with the CrowdTA framework.
Crowdsourcing allows outsourcing tasks to an anonymous group of individuals called crowdworkers, who are paid when completing the tasks. It has several benefits like cost-effectiveness and diversity of users. However, there are also some challenges mainly related to tasks' complexity and crowdworkers' related aspects such as their motivation. These challenges lead to lower quality in the results delivered by the workers. The purpose of this project is to design a gamification strategy for crowdsourcing using image annotation as a use case. In the end, a user-study will be conducted to validate whether the gamification strategy has an effect on the motivation of crowdworkers and the results' quality.