Refine
Year of publication
Document Type
- Master's Thesis (18)
- Part of Periodical (14)
- Doctoral Thesis (11)
- Bachelor Thesis (3)
- Study Thesis (2)
- Diploma Thesis (1)
- Habilitation (1)
Language
- English (50) (remove)
Keywords
- ontology (3)
- Linked Open Data (2)
- Maschinelles Lernen (2)
- mobile phone (2)
- multimedia metadata (2)
- 2019 European Parliament Election (1)
- API (1)
- Algolib (1)
- Analysis of social platform (1)
- Annotation (1)
Institute
- Institute for Web Science and Technologies (50) (remove)
Various best practices and principles guide an ontology engineer when modeling Linked Data. The choice of appropriate vocabularies is one essential aspect in the guidelines, as it leads to better interpretation, querying, and consumption of the data by Linked Data applications and users.
In this paper, we present the various types of support features for an ontology engineer to model a Linked Data dataset, discuss existing tools and services with respect to these support features, and propose LOVER: a novel approach to support the ontology engineer in modeling a Linked Data dataset. We demonstrate that none of the existing tools and services incorporate all types of supporting features and illustrate the concept of LOVER, which supports the engineer by recommending appropriate classes and properties from existing and actively used vocabularies. Hereby, the recommendations are made on the basis of an iterative multimodal search. LOVER uses different, orthogonal information sources for finding terms, e.g. based on a best string match or schema information on other datasets published in the Linked Open Data cloud. We describe LOVER's recommendation mechanism in general and illustrate it alongrna real-life example from the social sciences domain.
Ontologies play an important role in knowledge representation for sharing information and collaboratively developing knowledge bases. They are changed, adapted and reused in different applications and domains resulting in multiple versions of an ontology. The comparison of different versions and the analysis of changes at a higher level of abstraction may be insightful to understand the changes that were applied to an ontology. While there is existing work on detecting (syntactical) differences and changes in ontologies, there is still a need in analyzing ontology changes at a higher level of abstraction like ontology evolution or refactoring pattern. In our approach we start from a classification of model refactoring patterns found in software engineering for identifying such refactoring patterns in OWL ontologies using DL reasoning to recognize these patterns.
Ontologies are valuable tools for knowledge representation and important building blocks of the Semantic Web. They are not static and can change over time. Changing an ontology can be necessary for various reasons: the domain that is represented by an ontology can change or an ontology is reused and must be adapted to the new context. In addition, modeling errors could have been introduced into the ontology which must be found and removed. The non-triviality of the change process has led to the emerge of ontology change as an own field of research. The removal of knowledge from ontologies is an important aspect of this change process, because even the addition of new knowledge to an ontology potentially requires the removal of older, conflicting knowledge. Such a removal must be performed in a thought-out way. A naïve change of concepts within the ontology can easily remove other, unrelated knowledge or alter the semantics of concepts in an unintended way [2]. For these reasons, this thesis introduces a formal operator for the fine-grained retraction of knowledge from EL concepts which is partially based on the postulates for belief set contraction and belief base contraction [3, 4, 5] and the work of Suchanek et al. [6]. For this, a short introduction to ontologies and OWL 2 is given and the problem of ontology change is explained. It is then argued why a formal operator can support this process and why the Description Logic EL provides a good starting point for the development of such an operator. After this, a general introduction to Description Logic is given. This includes its history, an overview of its applications and common reasoning tasks in this logic. Following this, the logic EL is defined. In a next step, related work is examined and it is shown why the recovery postulate and the relevance postulate cannot be naïvely employed in the development of an operator that removes knowledge from EL concepts. Following this, the requirements to the operator are formulated and properties are given which are mainly based on the postulates for belief set and belief base contraction. Additional properties are developed which make up for the non-applicability of the recovery and relevance postulates. After this, a formal definition of the operator is given and it is shown that the operator is applicable to the task of a fine-grained removal of knowledge from EL concepts. In a next step, it is proven that the operator fulfills all the previously defined properties. It is then demonstrated how the operator can be combined with laconic justifications [7] to assist a human ontology editor by automatically removing unwanted consequences from an ontology. Building on this, a plugin for the ontology editor Protégé is introduced that is based on algorithms that were derived from the formal definition of the operator. The content of this work is then summarized and a final conclusion is drawn. The thesis closes with an outlook into possible future work.
We examine the systematic underrecognition of female scientists (Matilda effect) by exploring the citation network of papers published in the American Physical Society (APS) journals. Our analysis shows that articles written by men (first author, last author and dominant gender of authors) receive more citations than similar articles written by women (first author, last author and dominant gender of authors) after controlling for the journal of publication, year of publication and content of the publication. Statistical significance of the overlap between the lists of references was considered as the measure of similarity between articles in our analysis. In addition, we found that men are less likely to cite articles written by women and women are less likely to cite articles written by men. This pattern leads to receiving more citations by articles written by men than similar articles written by women because the majority of authors who published in APS journals are male (85%). We also observed Matilda effect reduces when articles are published in journals with the highest impact factors. In other words, people’s evaluation of articles published in these journals is not affected by the gender of authors significantly. Finally, we suggested a method that can be applied by editors in academic journals to reduce the evaluation bias to some extent. Editors can identify missing citations using our proposed method to complete bibliographies. This policy can reduce the evaluation bias because we observed papers written by female scholars (first author, last author, the dominant gender of authors) miss more citations than articles written by male scholars (first author, last author, the dominant gender of authors).
The distributed setting of RDF stores in the cloud poses many challenges. One such challenge is how the data placement on the compute nodes can be optimized to improve the query performance. To address this challenge, several evaluations in the literature have investigated the effects of existing data placement strategies on the query performance. A common drawback in theses evaluations is that it is unclear whether the observed behaviors were caused by the data placement strategies (if different RDF stores were evaluated as a whole) or reflect the behavior in distributed RDF stores (if cloud processing frameworks like Hadoop MapReduce are used for the evaluation). To overcome these limitations, this thesis develops a novel benchmarking methodology for data placement strategies that uses a data-placement-strategy-independent distributed RDF store to analyze the effect of the data placement strategies on query performance.
With this evaluation methodology the frequently used data placement strategies have been evaluated. This evaluation challenged the commonly held belief that data placement strategies that emphasize local computation, such as minimal edge-cut cover, lead to faster query executions. The results indicate that queries with a high workload may be executed faster on hash-based data placement strategies than on, e.g., minimal edge-cut covers. The analysis of the additional measurements indicates that vertical parallelization (i.e., a well-distributed workload) may be more important than horizontal containment (i.e., minimal data transport) for efficient query processing.
Moreover, to find a data placement strategy with a high vertical parallelization, the thesis tests the hypothesis that collocating small connected triple sets on the same compute node while balancing the amount of triples stored on the different compute nodes leads to a high vertical parallelization. Specifically, the thesis proposes two such data placement strategies. The first strategy called overpartitioned minimal edge-cut cover was found in the literature and the second strategy is the newly developed molecule hash cover. The evaluation revealed a balanced query workload and a high horizontal containment, which lead to a high vertical parallelization. As a result these strategies showed a better query performance than the frequently used data placement strategies.
We present the user-centered, iterative design of Mobile Facets, a mobile application for the faceted search and exploration of a large, multi-dimensional data set of social media on a touchscreen mobile phone. Mobile Facets provides retrieval of resources such as places, persons, organizations, and events from an integration of different open social media sources and professional content sources, namely Wikipedia, Eventful, Upcoming, geo-located Flickr photos, and GeoNames. The data is queried live from the data sources. Thus, in contrast to other approaches we do not know in advance the number and type of facets and data items the Mobile Facets application receives in a specific contextual situation. While developingrnMobile Facets, we have continuously evaluated it with a small group of fifive users. We have conducted a task-based, formative evaluation of the fifinal prototype with 12 subjects to show the applicability and usability of our approach for faceted search and exploration on a touchscreen mobile phone.
This thesis presents novel approaches for integrating context information into probabilistic models. Data from social media is typically associated with metadata, which includes context information such as timestamps, geographical coordinates or links to user profiles. Previous studies showed the benefits of using such context information in probabilistic models, e.g.\ improved predictive performance. In practice, probabilistic models which account for context information still play a minor role in data analysis. There are multiple reasons for this. Existing probabilistic models often are complex, the implementation is difficult, implementations are not publicly available, or the parameter estimation is computationally too expensive for large datasets. Additionally, existing models are typically created for a specific type of content and context and lack the flexibility to be applied to other data.
This thesis addresses these problems by introducing a general approach for modelling multiple, arbitrary context variables in probabilistic models and by providing efficient inference schemes and implementations.
In the first half of this thesis, the importance of context and the potential of context information for probabilistic modelling is shown theoretically and in practical examples. In the second half, the example of topic models is employed for introducing a novel approach to context modelling based on document clusters and adjacency relations in the context space. They can cope with areas of sparse observations and These models allow for the first time the efficient, explicit modelling of arbitrary context variables including cyclic and spherical context (such as temporal cycles or geographical coordinates). Using the novel three-level hierarchical multi-Dirichlet process presented in this thesis, the adjacency of ontext clusters can be exploited and multiple contexts can be modelled and weighted at the same time. Efficient inference schemes are derived which yield interpretable model parameters that allow analyse the relation between observations and context.
Data visualization is an effective way to explore data. It helps people to get a valuable insight of the data by placing it in a visual context. However, choosing a good chart without prior knowledge in the area is not a trivial job. Users have to manually explore all possible visualizations and decide upon ones that reflect relevant and desired trend in the data, are insightful and easy to decode, have a clear focus and appealing appearance. To address these challenges we developed a Tool for Automatic Generation of Good viSualizations using Scoring (TAG²S²). The approach tackles the problem of identifying an appropriate metric for judging visualizations as good or bad. It consists of two modules: visualization detection: given a data-set it creates a list of combination of data attributes for scoring and visualization ranking: scores each chart and decides which ones are good or bad. For the later, an utility metric of ten criteria was developed and each visualization detected in the first module is evaluated on these criteria. Only those visualizations that received enough scores are then presented to the user. Additionally to these data parameters, the tool considers user perception regarding the choice of visual encoding when selecting a visualization. To evaluate the utility of the metric and the importance of each criteria, test cases were developed, executed and the results presented.
In this thesis, I study the spectral characteristics of large dynamic networks and formulate the spectral evolution model. The spectral evolution model applies to networks that evolve over time, and describes their spectral decompositions such as the eigenvalue and singular value decomposition. The spectral evolution model states that over time, the eigenvalues of a network change while its eigenvectors stay approximately constant.
I validate the spectral evolution model empirically on over a hundred network datasets, and theoretically by showing that it generalizes arncertain number of known link prediction functions, including graph kernels, path counting methods, rank reduction and triangle closing. The collection of datasets I use contains 118 distinct network datasets. One dataset, the signed social network of the Slashdot Zoo, was specifically extracted during work on this thesis. I also show that the spectral evolution model can be understood as a generalization of the preferential attachment model, if we consider growth in latent dimensions of a network individually. As applications of the spectral evolution model, I introduce two new link prediction algorithms that can be used for recommender systems, search engines, collaborative filtering, rating prediction, link sign prediction and more.
The first link prediction algorithm reduces to a one-dimensional curve fitting problem from which a spectral transformation is learned. The second method uses extrapolation of eigenvalues to predict future eigenvalues. As special cases, I show that the spectral evolution model applies to directed, undirected, weighted, unweighted, signed and bipartite networks. For signed graphs, I introduce new applications of the Laplacian matrix for graph drawing, spectral clustering, and describe new Laplacian graph kernels. I also define the algebraic conflict, a measure of the conflict present in a signed graph based on the signed graph Laplacian. I describe the problem of link sign prediction spectrally, and introduce the signed resistance distance. For bipartite and directed graphs, I introduce the hyperbolic sine and odd Neumann kernels, which generalize the exponential and Neumann kernels for undirected unipartite graphs. I show that the problem of directed and bipartite link prediction are related by the fact that both can be solved by considering spectral evolution in the singular value decomposition.
Next word prediction is the task of suggesting the most probable word a user will type next. Current approaches are based on the empirical analysis of corpora (large text files) resulting in probability distributions over the different sequences that occur in the corpus. The resulting language models are then used for predicting the most likely next word. State-of-the-art language models are based on n-grams and use smoothing algorithms like modified Kneser-Ney smoothing in order to reduce the data sparsity by adjusting the probability distribution of unseen sequences. Previous research has shown that building word pairs with different distances by inserting wildcard words into the sequences can result in better predictions by further reducing data sparsity. The aim of this thesis is to formalize this novel approach and implement it by also including modified Kneser-Ney smoothing.