Filtern
Erscheinungsjahr
Dokumenttyp
- Masterarbeit (19)
- Ausgabe (Heft) zu einer Zeitschrift (15)
- Dissertation (11)
- Studienarbeit (5)
- Bachelorarbeit (3)
- Diplomarbeit (3)
- Habilitation (1)
Schlagworte
- Semantic Web (3)
- ontology (3)
- Linked Open Data (2)
- Maschinelles Lernen (2)
- OWL (2)
- OWL <Informatik> (2)
- Ontology (2)
- RDF <Informatik> (2)
- SPARQL (2)
- mobile phone (2)
Institut
- Institute for Web Science and Technologies (57) (entfernen)
Most social media platforms allow users to freely express their opinions, feelings, and beliefs. However, in recent years the growing propagation of hate speech, offensive language, racism and sexism on the social media outlets have drawn attention from individuals, companies, and researchers. Today, sexism both online and offline with different forms, including blatant, covert, and subtle lan- guage, is a common phenomenon in society. A notable amount of work has been done over identifying sexist content and computationally detecting sexism which exists online. Although previous efforts have mostly used peoples’ activities on social media platforms such as Twitter as a public and helpful source for collecting data, they neglect the fact that the method of gathering sexist tweets could be biased towards the initial search terms. Moreover, some forms of sexism could be missed since some tweets which contain offensive language could be misclassified as hate speech. Further, in existing hate speech corpora, sexist tweets mostly express hostile sexism, and to some degree, the other forms of sexism which also appear online was disregarded. Besides, the creation of labeled datasets with manual exertion, relying on users to report offensive comments with a tremendous effort by human annotators is not only a costly and time-consuming process, but it also raises the risk of involving discrimination under biased judgment.
This thesis generates a novel sexist and non-sexist dataset which is constructed via "UnSexistifyIt", an online web-based game that incentivizes the players to make minimal modifications to a sexist statement with the goal of turning it into a non-sexist statement and convincing other players that the modified statement is non-sexist. The game applies the methodology of "Game With A Purpose" to generate data as a side-effect of playing the game and also employs the gamification and crowdsourcing techniques to enhance non-game contexts. When voluntary participants play the game, they help to produce non-sexist statements which can reduce the cost of generating new corpus. This work explores how diverse individual beliefs concerning sexism are. Further, the result of this work highlights the impact of various linguistic features and content attributes regarding sexist language detection. Finally, this thesis could help to expand our understanding regarding the syntactic and semantic structure of sexist and non-sexist content and also provides insights to build a probabilistic classifier for single sentences into sexist or non-sexist classes and lastly find a potential ground truth for such a classifier.
Knowledge-based authentication methods are vulnerable to Shoulder surfing phenomenon.
The widespread usage of these methods and not addressing the limitations it has could result in the user’s information to be compromised. User authentication method ought to be effortless to use and efficient, nevertheless secure.
The problem that we face concerning the security of PIN (Personal Identification Number) or password entry is shoulder surfing, in which a direct or indirect malicious observer could identify the user sensitive information. To tackle this issue we present TouchGaze which combines gaze signals and touch capabilities, as an input method for entering user’s credentials. Gaze signals will be primarily used to enhance targeting and touch for selecting. In this work, we have designed three different PIN entry method which they all have similar interfaces. For the evaluation, these methods were compared based on efficiency, accuracy, and usability. The results uncovered that despite the fact that gaze-based methods require extra time for the user to get familiar with yet it is considered more secure. In regards to efficiency, it has the similar error margin to the traditional PIN entry methods.
The novel mobile application csxPOI (short for: collaborative, semantic, and context-aware points-of-interest) enables its users to collaboratively create, share, and modify semantic points of interest (POI). Semantic POIs describe geographic places with explicit semantic properties of a collaboratively created ontology. As the ontology includes multiple subclassiffcations and instantiations and as it links to DBpedia, the richness of annotation goes far beyond mere textual annotations such as tags. With the intuitive interface of csxPOI, users can easily create, delete, and modify their POIs and those shared by others. Thereby, the users adapt the structure of the ontology underlying the semantic annotations of the POIs. Data mining techniques are employed to cluster and thus improve the quality of the collaboratively created POIs. The semantic POIs and collaborative POI ontology are published as Linked Open Data.
Entwicklung eines generischen Sesame-Sails für die Abbildung von SPARQL-Anfragen auf Webservices
(2010)
Diese Arbeit soll eine Möglichkeit aufzeigen, aufbauend auf dem Sesame Framework Datenbestände von nicht-semantischen Web-Diensten im Sinne des Semantic Web auszuwerten. Konkret wird ein Sail (Webservice-Sail) entwickelt, das einen solchen Web-Dienst wie eine RDF-Quelle abfragen kann, indem es SPARQL-Ausdrücke in Methodenaufrufe des Dienstes übersetzt und deren Ergebnisse entsprechend auswertet und zurückgibt. Um eine möglichst große Anzahl von Webservices abdecken zu können, muss die Lösung entsprechend generisch gehalten sein. Das bedeutet aber insbesondere auch, dass das Sail auf die Modalitäten konkreter Services eingestellt werden muss. Es muss also auch eine geeignete Konfigurationsrepräsentation gefunden werden, um eine möglichst gute Unterstützung eines zu verwendenden Web-Dienstes durch das Webservice-Sail zu gewährleisten. Die Entwicklung einer solchen Repräsentation ist damit auch Bestandteil dieser Arbeit.
Belief revision is the subarea of knowledge representation which studies the dynamics of epistemic states of an agent. In the classical AGM approach, contraction, as part of the belief revision, deals with the removal of beliefs in knowledge bases. This master's thesis presents the study and the implementation of concept contraction in the Description Logic EL. Concept contraction deals with the following situation. Given two concept C and D, assuming that C is subsumed by D, how can concept C be changed so that it is not subsumed by D anymore, but is as similar as possible to C? This approach of belief change is different from other related work because it deals with contraction in the level of concepts and not T-Boxes and A-Boxes in general. The main contribution of the thesis is the implementation of the concept contraction. The implementation provides insight into the complexity of contraction in EL, which is tractable since the main inference task in EL is also tractable. The implementation consists of the design of five algorithms that are necessary for concept contraction. The algorithms are described, illustrated with examples, and analyzed in terms of time complexity. Furthermore, we propose an new approach for a selection function, adapt for the concept contraction. The selection function uses metadata about the concepts in order to select the best from an input set. The metadata is modeled in a framework that we have designed, based on standard metadata frameworks. As an important part of the concept contraction, the selection function is responsible for selecting the best concepts that are as similar as possible to concept C. Lastly, we have successfully implemented the concept contraction in Python, and the results are promising.
Tagging-Systeme sind faszinierende dynamische Systeme in denen Benutzer kollaborativ Ressourcen mit sogenannten Tags indexieren. Um das volle Potential von Tagging-Systemen nutzen zu können ist es wichtig zu verstehen, wie sich das Verhalten der einzelnen Benutzer auf die Eigenschaften des Gesamtsystems auswirkt. In der vorliegenden Arbeit wird das Epistemic Dynamic Model präsentiert. Es schlägt eine Brücke zwischen dem Benutzerverhalten und den Systemeigenschaften. Das Modell basiert auf der Annahme, dass der Einfluss des gemeinsamen Hintergrundwissens der Benutzer und der Imitation von Tag-Vorschlägen ausreicht, um die Entstehung der Häufigkeitsverteilungen der Tags und des Wachstums des Vokabulars zu erklären. Diese beiden Eigenschaften eines Tagging-Systems hängen eng mit der Entstehung eines gemeinsamen Vokabulars der Benutzer zusammen. Mit Hilfe des Epistemic Dynamic Models zeigen wir, dass die generelle Ausprägung der Tag-Häufigkeitsverteilungen und des Wachstums des Vokabulars ihren Ursprung in dem gemeinsamen Hintergrundwissen der Benutzer haben. Tag-Vorschläge können dann dazu genutzt werden, um gezielt diese generelle Ausprägung zu beeinflussen. In der vorliegenden Arbeit untersuchen wir hauptsächlich den Einfluss der von Vorschlägen populärer Tags ausgeht. Populäre Tags sorgen für einen Feedback-Mechanismus zwischen den Vokabularen der einzelnen Benutzer, der die Inter-Indexer Konsistenz der Tag-Zuweisungen erhöht. Wie wird aber dadurch die Indexierungsqualität in Tagging-Systemen beeinflusst? Zur Klärung dieser Frage untersuchen wir eine Methode zur Messung der Inter-Ressourcen Konsistenz der Tag-Zuweisungen. Die Inter-Ressourcen Konsistenz korreliert positiv mit der Indexierungsqualität, und mit der Trefferquote und der Genauigkeit von Suchanfragen an das System. Sie misst inwieweit die Tag-Vektoren die durch Benutzer wahrgenommene Ähnlichkeit der jeweiligen Ressourcen widerspiegeln. Wir legen mit Hilfe unseres Modell dar, und zeigen es auch mit Hilfe eines Benutzerexperiments, dass populäre Tags zu einer verringerten Inter-Ressourcen Konsistenz führen. Des Weiteren zeigen wir, dass die Inter-Ressourcen Konsistenz erhöht wird, wenn dem Benutzer das eigene, bisher genutzte Vokabular vorgeschlagen wird. Unsere Methode zur Messung der Inter-Ressourcen Konsistenz ergänzt bestehende Evaluationsmaße für Tag-Vorschlags-Algorithmen um den Aspekt der Indexierungsqualität.
Navigation is a natural way to explore and discover content in a digital environment. Hence, providers of online information systems such as Wikipedia---a free online encyclopedia---are interested in providing navigational support to their users. To this end, an essential task approached in this thesis is the analysis and modeling of navigational user behavior in information networks with the goal of paving the way for the improvement and maintenance of web-based systems. Using large-scale log data from Wikipedia, this thesis first studies information access by contrasting search and navigation as the two main information access paradigms on the Web. Second, this thesis validates and builds upon existing navigational hypotheses to introduce an adaptation of the well-known PageRank algorithm. This adaptation is an improvement of the standard PageRank random surfer navigation model that results in a more "reasonable surfer" by accounting for the visual position of links, the information network regions they lead to, and the textual similarity between the link source and target articles. Finally, using agent-based simulations, this thesis compares user models that have a different knowledge of the network topology in order to investigate the amount and type of network topological information needed for efficient navigation. An evaluation of agents' success on four different networks reveals that in order to navigate efficiently, users require only a small amount of high-quality knowledge of the network topology. Aside from the direct benefits to content ranking provided by the "reasonable surfer" version of PageRank, the empirical insights presented in this thesis may also have an impact on system design decisions and Wikipedia editor guidelines, i.e., for link placement and webpage layout.
The Web contains some extremely valuable information; however, often poor quality, inaccurate, irrelevant or fraudulent information can also be found. With the increasing amount of data available, it is becoming more and more difficult to distinguish truth from speculation on the Web. One of the most, if not the most, important criterion used to evaluate data credibility is the information source, i.e., the data origin. Trust in the information source is a valuable currency users have to evaluate such data. Data popularity, recency (or the time of validity), reliability, or vagueness ascribed to the data may also help users to judge the validity and appropriateness of information sources. We call this knowledge derived from the data the provenance of the data. Provenance is an important aspect of the Web. It is essential in identifying the suitability, veracity, and reliability of information, and in deciding whether information is to be trusted, reused, or even integrated with other information sources. Therefore, models and frameworks for representing, managing, and using provenance in the realm of Semantic Web technologies and applications are critically required. This thesis highlights the benefits of the use of provenance in different Web applications and scenarios. In particular, it presents management frameworks for querying and reasoning in the Semantic Web with provenance, and presents a collection of Semantic Web tools that explore provenance information when ranking and updating caches of Web data. To begin, this thesis discusses a highly exible and generic approach to the treatment of provenance when querying RDF datasets. The approach re-uses existing RDF modeling possibilities in order to represent provenance. It extends SPARQL query processing in such a way that given a SPARQL query for data, one may request provenance without modifying it. The use of provenance within SPARQL queries helps users to understand how RDF facts arederived, i.e., it describes the data and the operations used to produce the derived facts. Turning to more expressive Semantic Web data models, an optimized algorithm for reasoning and debugging OWL ontologies with provenance is presented. Typical reasoning tasks over an expressive Description Logic (e.g., using tableau methods to perform consistency checking, instance checking, satisfiability checking, and so on) are in the worst case doubly exponential, and in practice are often likewise very expensive. With the algorithm described in this thesis, however, one can efficiently reason in OWL ontologies with provenance, i.e., provenance is efficiently combined and propagated within the reasoning process. Users can use the derived provenance information to judge the reliability of inferences and to find errors in the ontology. Next, this thesis tackles the problem of providing to Web users the right content at the right time. The challenge is to efficiently rank a stream of messages based on user preferences. Provenance is used to represent preferences, i.e., the user defines his preferences over the messages' popularity, recency, etc. This information is then aggregated to obtain a joint ranking. The aggregation problem is related to the problem of preference aggregation in Social Choice Theory. The traditional problem formulation of preference aggregation assumes a I fixed set of preference orders and a fixed set of domain elements (e.g. messages). This work, however, investigates how an aggregated preference order has to be updated when the domain is dynamic, i.e., the aggregation approach ranks messages 'on the y' as the message passes through the system. Consequently, this thesis presents computational approaches for online preference aggregation that handle the dynamic setting more efficiently than standard ones. Lastly, this thesis addresses the scenario of caching data from the Linked Open Data (LOD) cloud. Data on the LOD cloud changes frequently and applications relying on that data - by pre-fetching data from the Web and storing local copies of it in a cache - need to continually update their caches. In order to make best use of the resources (e.g., network bandwidth for fetching data, and computation time) available, it is vital to choose a good strategy to know when to fetch data from which data source. A strategy to cope with data changes is to check for provenance. Provenance information delivered by LOD sources can denote when the resource on the Web has been changed last. Linked Data applications can benefit from this piece of information since simply checking on it may help users decide which sources need to be updated. For this purpose, this work describes an investigation of the availability and reliability of provenance information in the Linked Data sources. Another strategy for capturing data changes is to exploit provenance in a time-dependent function. Such a function should measure the frequency of the changes of LOD sources. This work describes, therefore, an approach to the analysis of data dynamics, i.e., the analysis of the change behavior of Linked Data sources over time, followed by the investigation of different scheduling update strategies to keep local LOD caches up-to-date. This thesis aims to prove the importance and benefits of the use of provenance in different Web applications and scenarios. The exibility of the approaches presented, combined with their high scalability, make this thesis a possible building block for the Semantic Web proof layer cake - the layer of provenance knowledge.
With the Multimedia Metadata Ontology (M3O), we have developed a sophisticated model for representing among others the annotation, decomposition, and provenance of multimedia metadata. The goal of the M3O is to integrate the existing metadata standards and metadata formats rather than replacing them. To this end, the M3O provides a scaffold needed to represent multimedia metadata. Being an abstract model for multimedia metadata, it is not straightforward how to use and specialize the M3O for concrete application requirements and existing metadata formats and metadata standards. In this paper, we present a step-by-step alignment method describing how to integrate and leverage existing multimedia metadata standards and metadata formats in the M3O in order to use them in a concrete application. We demonstrate our approach by integrating three existing metadata models: the Core Ontology on Multimedia (COMM), which is a formalization of the multimedia metadata standard MPEG-7, the Ontology for Media Resource of the W3C, and the widely known industry standard EXIF for image metadata
Expert-driven business process management is an established means for improving efficiency of organizational knowledge work. Implicit procedural knowledge in the organization is made explicit by defining processes. This approach is not applicable to individual knowledge work due to its high complexity and variability. However, without explicitly described processes there is no analysis and efficient communication of best practices of individual knowledge work within the organization. In addition, the activities of the individual knowledge work cannot be synchronized with the activities in the organizational knowledge work.rnrnSolution to this problem is the semantic integration of individual knowledgernwork and organizational knowledge work by means of the patternbased core ontology strukt. The ontology allows for defining and managing the dynamic tasks of individual knowledge work in a formal way and to synchronize them with organizational business processes. Using the strukt ontology, we have implemented a prototype application for knowledge workers and have evaluated it at the use case of an architectural fifirm conducting construction projects.
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.
In this article we analyze the privacy aspects of a mobile sensor application used for recording urban travel patterns as part of a travel-survey service. This service has been developed and field-tested within the Live+Gov EU Project. The privacy analysis follows a structured approach established in. Eight privacy recommendations are derived, and have already led to corresponding enhancements of the travel-survey service.
Ontologien sind wichtige Werkzeuge zur Wissensrepräsentation und elementare Bausteine des Semantic Web. Sie sind jedoch nicht statisch und können sich über die Zeit verändern. Die Gründe hierfür sind vielfältig: Konzepte innerhalb einer Ontologie können fehlerhaft modelliert worden sein, die von der Ontologie repräsentierte Domäne kann sich verändern oder eine Ontologie kann wiederverwendet werden und muss an den neuen Kontext angepasst oder mit bestehenden Ontologien verbunden werden. Die Schwierigkeit dieses Prozesses hat zur Entstehung des Forschungsfeldes der Ontology Change geführt. Das Entfernen von Wissen aus Ontologien ist ein wichtiger Aspekt dieses Änderungsprozesses, da selbst das Hinzufügen neuen Wissens zu einer Ontologie das Entfernen bestehenden Wissens notwendig machen kann, falls dieses mit den neuen Vorstellungen in Konflikt steht. Dieses Entfernen muss jedoch wohldurchdacht sein, da das Ändern bestehender Konzepte leicht zu viel Wissen aus der Ontologie entfernen oder die semantische Bedeutung der Konzepte auf eine potenziell unerwartete Weise verändern kann. In dieser Arbeit wird daher ein formaler Operator zum präzisen Entfernen von Wissen aus Konzepten vorgestellt. Dieser basiert auf der Beschreibungslogik EL und baut partiell auf den Postulaten für Belief Set und Belief Base Contraction sowie der Arbeit von Suchanek et al. auf. Hierfür wird zunächst ein Einstieg in das Thema Ontologien und die Ontologiesprache OWL 2 gegeben und das Problemfeld der Ontology Change wird erläutert. Es wird dann gezeigt, wie ein formaler Operator diesen Prozess unterstützen kann und weshalb die Beschreibungslogik EL einen guten Ausgangspunkt für die Entwicklung eines solchen Operators darstellt. Anschließend wird ein Einblick in das Feld der Beschreibungslogiken gegeben. Hierfür wird die Geschichte der Beschreibungslogik kurz umrissen, Anwendungsgebiete werden genannt und es werden Standardprobleme in dieser Logik erläutert. In diesem Zusammenhang wird die Beschreibungslogik EL formal eingeführt. In einem nächsten Schritt werden verwandte Arbeiten untersucht und es wird gezeigt, warum das Recovery- und Relevance-Postulat für das Entfernen von Wissen aus Konzepten nicht unmittelbar anwendbar ist. Die hier gewonnenen Erkenntnisse werden anschließend dazu genutzt, die Anforderungen an den Operator zu formalisieren. Diese basieren hauptsächlich auf den Postulaten für Belief Set und Belief Base Contraction. Zusätzlich werden weitere Eigenschaften formuliert welche den Verlust des Recovery- bzw. Relevance-Postulates ausgleichen sollen. In einem nächsten Schritt wird der Operator definiert und es wird gezeigt, dass diese Definition das präzise Entfernen von Wissen aus EL-Konzepten gestattet. Mittels formaler Beweise wird zudem gezeigt, dass diese Definition alle zuvor aufgestellten Anforderungen erfüllt. In einem weiteren Beispiel wird dargestellt, wie der Operator in Verbindung mit sogenannten Laconic Justifications verwendet werden kann, um einen menschlichen Ontology-Editor durch das automatisierte Entfernen von unerwünschten Konsequenzen aus der Ontologie zu unterstützen. Aufbauend auf Algorithmen, welche aus der formalen Definition des Operators abgeleitet wurden, wird ein Plugin zum Entfernen von Wissen aus Ontologien für den Ontology-Editor Protégé vorgestellt. Anschließend werden die bisherigen Erkenntnisse zusammengefasst und es wird ein Fazit gezogen. Die Arbeit schließt mit einem Ausblick über mögliche zukünftige Forschung.
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.
In dieser Arbeit wird das MobileFacets System präsentiert, dass ein bequemes facettiertes Browsen und Suchen von semantischen Daten auf einem mobilen Endgerät ermöglicht. Anwender bekommen in Abhängigkeit ihres lokalen Ortskontextes, weitreichende Informationen wie Orte, Personen, Organisationen oder Events dargeboten. Basierend auf der Theorie von Facetten, wird das facettierte Browsen zur Erkundung von strukturierten Datensätzen anhand einer Client Anwendung realisiert. Die Anwendung bedient sich dabei eines lokalen Servers, der für Anfragen der Clients, die Anbindung an externe Datenquellen und die Aufbereitung der strukturierten Daten zuständig ist.
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.