Contributions in Merlin

As part of the 6th Lifelog Search Challenge, this paper presents an approach to arrange Lifelog data in a multi-modal knowledge graph based on cluster hierarchies. We use multiple sequence clustering approaches to address the multi-modal nature of Lifelogs in relation to temporal, spatial, and visual factors. The resulting clusters, along with semantic metadata captions and augmentations based on OpenCLIP, provide for the semantic structure of a graph including all Lifelogs as entries. Textual queries on this hierarchical graph can be expressed to retrieve individual Lifelogs, as well as clusters of Lifelogs.

The war in Ukraine and the following refugee crisis have recently again highlighted the need for effective refugee management across European countries. Refugee management contemporarily mostly relies on top-down management approaches by governments. These often lead to suboptimal policies for refugees and highlight a need to better identify and integrate refugee needs into management. Here, we show that modern applications of Natural Language Processing (NLP) allow for the effective analysis of large text corpora linked to refugee needs, making it possible to complement top-down approaches with bottom-up knowledge centered around the current needs of the refugee population. By following a Design Science Research Methodology, we utilize 58 semi-structured stakeholder interviews within Switzerland to develop design requirements for NLP applications for refugee management. Based on the design requirements, we developed R2G – “Refugees to Government”, an application based on state-of-the-art topic modeling to identify refugee needs bottom-up through Telegram data. We evaluate R2G with a dedicated workshop held with stakeholders from the public sector and civil society. Thus, we contribute to the ongoing discourse on how to design refugee management applications and showcase how topic modeling can be utilized for data-driven governance during refugee crises.

Multi-task learning (MTL) aims at achieving a better model by leveraging data and knowledge from multiple tasks. However, MTL does not always work – sometimes negative transfer occurs between tasks, especially when aggregating loosely related skills, leaving it an open question when MTL works. Previous studies show that MTL performance can be improved by algorithmic tricks. However, what tasks and skills should be included is less well explored. In this work, we conduct a case study in Financial NLP where multiple datasets exist for skills relevant to the domain, such as numeric reasoning and sentiment analysis. Due to the task difficulty and data scarcity in the Financial NLP domain, we explore when aggregating such diverse skills from multiple datasets with MTL can work. Our findings suggest that the key to MTL success lies in skill diversity, relatedness between tasks, and choice of aggregation size and shared capacity. Specifically, MTL works well when tasks are diverse but related, and when the size of the task aggregation and the shared capacity of the model are balanced to avoid overwhelming certain tasks.

To transition to a green economy, environmental claims made by companies must be reliable, comparable, and verifiable. To analyze such claims at scale, automated methods are needed to detect them in the first place. However, there exist no datasets or models for this. Thus, this paper introduces the task of environmental claim detection. To accompany the task, we release an expert-annotated dataset and models trained on this dataset. We preview one potential application of such models: We detect environmental claims made in quarterly earning calls and find that the number of environmental claims has steadily increased since the Paris Agreement in 2015.

Research papers on ESG themes - positive and sceptical - dominate recent downloads from the Social Science Research Network website.

The rise of opinion programs has transformed television news. Because they present anchors’ subjective commentary and analysis, opinion programs often convey conflicting narratives about reality. We experimentally document that people across the ideological spectrum turn to opinion programs over “straight news”, even when provided large incentives to learn objective facts. We then examine the consequences of diverging narratives between opinion programs in a high-stakes setting: the early stages of the COVID-19 pandemic in the US. We find stark differences in the adoption of preventative behaviours among viewers of the two most popular opinion programs, both on the same network, which adopted opposing narratives about the threat posed by the COVID-19 pandemic. We then show that areas with greater relative viewership of the program downplaying the threat experienced a greater number of COVID-19 cases and deaths. Our evidence suggests that opinion programs may distort important beliefs and behaviours.

Traditional explanatory resources, such as user manuals and textbooks, often contain content that may not cater to the diverse backgrounds and information needs of users. Yet, developing intuitive, user-centered methods to effectively explain complex or large amounts of information is still an open research challenge. In this paper we present ExplanatoryGPT, an approach we devised and implemented to transform textual documents into interactive, intelligent resources, capable of offering dynamic, personalized explanations. Our approach uses state-of-the-art question-answering technology to generate on-demand, expandable explanations, with the aim of allowing readers to efficiently navigate and comprehend static materials. ExplanatoryGPT integrates ChatGPT, a state-of-the-art language model, with Achinstein’s philosophical theory of explanations. By combining question generation and answer retrieval algorithms with ChatGPT, our method generates interactive, user-centered explanations, while mitigating common issues associated with ChatGPT, such as hallucinations and memory shortcomings. To showcase the effectiveness of our Explanatory AI, we conducted tests using a variety of sources, including a legal textbook and documentation of some health and financial software. Specifically, we provide several examples that illustrate how ExplanatoryGPT excels over ChatGPT in generating more precise explanations, accomplished through thoughtful macro-planning of explanation content. Notably, our approach also avoids the need to provide the entire context of the explanation as a prompt to ChatGPT, a process that is often not feasible due to common memory constraints.

This thesis aims to address a key challenge in time-series data analysis - efficiently identifying correlations between various data streams. As the prominence of sensor networks rises, the analysis of time-series data has become increasingly crucial. The insights derived from these data streams hold significant intelligence, but extracting them efficiently remains a complex task. This thesis brings into focus the dimensionality-reduction filter-and-refine techniques, designed to expedite the process of identifying correlations. Despite their utility, these techniques lack a comprehensive comparative analysis over streaming systems, and this thesis seeks to fill this gap. The core objective is to implement these techniques within a streaming platform, enabling benchmarking under realistic conditions. The thesis is divided into several chapters that provide a comprehensive overview of the problem, delve into the dimensionality reduction algorithms, and discuss their implementation within a streaming system. Particular emphasis is placed on the Filter-and-Refine Algorithm on the Kafka Streaming Platform. Two distinct design approaches, one based on Kafka Streams and another simpler Producer-Consumer design, are implemented, compared, and evaluated. The thesis culminates with an exhaustive series of experiments assessing the performance of the implemented algorithms. The ultimate goal is to provide a framework that not only implements the techniques but also evaluates their performance, aiming to contribute to the field of time-series data analysis.

The number of people suffering from chronic diseases worldwide is rapidly increasing, leading to rising costs and decreased quality of life. Lifestyle changes are an effective measure in the treatment of chronic diseases such as obesity. However, treatment adherence is consistently low. This thesis aims to analyse how the initial phases of the Digital Companion influence patient empowerment as well as self-perceived and actual adherence to the treatment plan. We analyse 18 patient interviews conducted during the field study of the Digital Companion. The results indicate that the structural empowerment provided by the Health Care Professional (HCP) together with the app have a positive effect on patients’ psychological empowerment. Patients report high levels of self-perceived empowerment and predicted adherence to the treatment plan which they planned with their HCP. The evaluation of patient-generated data recorded in the app shows a very high average actual adherence of 80%.

Ultra-Wideband (UWB) technology has gained significant popularity in indoor localization applications. These applications often generate vast amounts of personal information, increasing the need to ensure compliance with privacy-preserving principles to safeguard user data. In this thesis, the privacy-preserving characteristics of several UWB localization architectures were analyzed. Firstly, UWB localization architectures were examined based on their privacy-preserving characteristics. Subsequently, two versions of a time difference of arrival (tdoa) localization system were implemented, including privacy best practices provided by the IEEE 802.15.4 standard during the implementation process. Additionally, the privacy-preserving characteristics of the implemented UWB localization systems were evaluated with the help of a privacy criteria catalog based on COPri V.2 ontology. This thesis found that the localization system employing a passively listening tag fulfills seven out of eight privacy criteria. In contrast, the system where the tag actively sends out UWB signals only fulfilled three out of eight criteria in its minimal version. However, the privacy-preserving characteristics of the active system could be greatly improved by using tools such as dynamic addressing, encrypting packages containing personal information, using a message integrity code (MIC), and using a scrambled time sequence (STS). Finally, the limitations of the current systems' implementations are addressed which provides directions for future research.