05 Fakultät Informatik, Elektrotechnik und Informationstechnik
Permanent URI for this collectionhttps://elib.uni-stuttgart.de/handle/11682/6
Browse
2 results
Search Results
Item Open Access Geometric relational embeddings(2024) Xiong, Bo; Staab, Steffen (Prof. Dr.)In classical AI, symbolic knowledge is typically represented as relational data within a graph-structured framework, a.k.a., relational knowledge bases (KBs). Relational KBs suffer from incompleteness and numerous efforts have been dedicated to KB completion. One prevalent approach involves mapping relational data into continuous representations within a low-dimensional vector space, referred to as relational representation learning. This facilitates the preservation of relational structures, allowing for effective inference of missing knowledge from the embedding space. Nevertheless, existing methods employ pure-vector embeddings and map each relational object, such as entities, concepts, or relations, as a simple point in a vector space (typically Euclidean. While these pure-vector embeddings are simple and adept at capturing object similarities, they fall short in capturing various discrete and symbolic properties inherent in relational data. This thesis surpasses conventional vector embeddings by embracing geometric embeddings to more effectively capture the relational structures and underlying discrete semantics of relational data. Geometric embeddings map data objects as geometric elements, such as points in hyperbolic space with constant negative curvature or convex regions (e.g., boxes, disks) in Euclidean vector space, offering superior modeling of discrete properties present in relational data. Specifically, this dissertation introduces various geometric relational embedding models capable of capturing: 1) complex structured patterns like hierarchies and cycles in networks and knowledge graphs; 2) intricate relational/logical patterns in knowledge graphs; 3) logical structures in ontologies and logical constraints applicable for constraining machine learning model outputs; and 4) high-order complex relationships between entities and relations. Our results obtained from benchmark and real-world datasets demonstrate the efficacy of geometric relational embeddings in adeptly capturing these discrete, symbolic, and structured properties inherent in relational data, which leads to performance improvements over various relational reasoning tasks.Item Open Access Spatiotemporal fusion of nonverbal voice & eye gaze for human-computer interactions(2025) Hedeshy, Ramin; Staab, Steffen (Prof. Dr.)This dissertation explores the novel concept of hands-free interaction through nonverbal voice expressions (NVVEs) and eye gaze, with a particular focus on core aspects of human computer interaction, text entry and point-and-click. The benefits of this research are principally notable for individuals dealing with physical disabilities or challenges related to speech, offering them a more intuitive and inclusive way to interact with the digital world. Moreover, it could also be applicable in settings where dictation suffers from poor voice recognition or where spoken words and sentences jeopardize privacy or confidentiality. We introduce two innovative hands-free text entry methods by analyzing temporally constrained gaze paths accompanied by simple touch or auditory signals from nonverbal vocalizations such as humming. Experimental evaluations demonstrate that these methods outperform the traditional eye gaze typing techniques in terms of speed, accuracy, and overall user satisfaction. The dissertation further extends this approach to create an intuitive control mechanism for point-and-click systems within a gaming application using NVVEs synchronized with eye gaze. A representative interface "All Birds Must Fly" was developed to validate the technique among people without disabilities as well as those who are physically challenged; results indicate not only effective game environment control but also enhanced engagement level compared to conventional mouse and keyboard setup. To overcome the limitations of using NVVEs such as limited availability of suitable training data and computational methods for classifying such expressions in noisy environment that has constrained the exploration of this technique to a focus on simple binary inputs, a dataset was collected which has been made publicly accessible. We provide a Convolutional Neural Network (CNN) model with a test accuracy of 96.6% in a 5-fold cross-validation that can classify 6 type of expressions.