Refine
Has Fulltext
- yes (159) (remove)
Year of publication
- 2022 (159) (remove)
Document Type
- Doctoral Thesis (159) (remove)
Is part of the Bibliography
- yes (159) (remove)
Keywords
- Klimawandel (5)
- climate change (5)
- Digitalisierung (3)
- Modellierung (3)
- Röntgenspektroskopie (3)
- modelling (3)
- Adipositas (2)
- Arbeitszufriedenheit (2)
- Bewegungsökologie (2)
- Bundeswehr (2)
Institute
- Institut für Biochemie und Biologie (27)
- Extern (26)
- Institut für Physik und Astronomie (20)
- Institut für Chemie (19)
- Institut für Geowissenschaften (17)
- Hasso-Plattner-Institut für Digital Engineering GmbH (14)
- Institut für Umweltwissenschaften und Geographie (7)
- Wirtschaftswissenschaften (7)
- Institut für Ernährungswissenschaft (5)
- Institut für Mathematik (5)
The main goal of this dissertation is to experimentally investigate how focus is realised, perceived, and processed by native Turkish speakers, independent of preconceived notions of positional restrictions. Crucially, there are various issues and scientific debates surrounding focus in the Turkish language in the existing literature (chapter 1). It is argued in this dissertation that two factors led to the stagnant literature on focus in Turkish: the lack of clearly defined, modern understandings of information structure and its fundamental notion of focus, and the ongoing and ill-defined debate surrounding the question of whether there is an immediately preverbal focus position in Turkish. These issues gave rise to specific research questions addressed across this dissertation. Specifically, we were interested in how the focus dimensions such as focus size (comparing narrow constituent and broad sentence focus), focus target (comparing narrow subject and narrow object focus), and focus type (comparing new-information and contrastive focus) affect Turkish focus realisation and, in turn, focus comprehension when speakers are provided syntactic freedom to position focus as they see fit.
To provide data on these core goals, we presented three behavioural experiments based on a systematic framework of information structure and its notions (chapter 2): (i) a production task with trigger wh-questions and contextual animations manipulated to elicit the focus dimensions of interest (chapter 3), (ii) a timed acceptability judgment task in listening to the recorded answers in our production task (chapter 4), and (iii) a self-paced reading task to gather on-line processing data (chapter 5).
Based on the results of the conducted experiments, multiple conclusions are made in this dissertation (chapter 6). Firstly, this dissertation demonstrated empirically that there is no focus position in Turkish, neither in the sense of a strict focus position language nor as a focally loaded position facilitating focus perception and/or processing. While focus is, in fact, syntactically variable in the Turkish preverbal area, this is a consequence of movement triggered by other IS aspects like topicalisation and backgrounding, and the observational markedness of narrow subject focus compared to narrow object focus. As for focus type in Turkish, this dimension is not associated with word order in production, perception, or processing. Significant acoustic correlates of focus size (broad sentence focus vs narrow constituent focus) and focus target (narrow subject focus vs narrow object focus) were observed in fundamental frequency and intensity, representing focal boost, (postfocal) deaccentuation, and the presence or absence of a phrase-final rise in the prenucleus, while the perceivability of these effects remains to be investigated. In contrast, no acoustic correlates of focus type in simple, three-word transitive structures were observed, with focus types being interchangeable in mismatched question-answer pairs. Overall, the findings of this dissertation highlight the need for experimental investigations regarding focus in Turkish, as theoretical predictions do not necessarily align with experimental data. As such, the fallacy of implying causation from correlation should be strictly kept in mind, especially when constructions coincide with canonical structures, such as the immediately preverbal position in narrow object foci. Finally, numerous open questions remain to be explored, especially as focus and word order in Turkish are multifaceted. As shown, givenness is a confounding factor when investigating focus types, while thematic role assignment potentially confounds word order preferences. Further research based on established, modern information structure frameworks is needed, with chapter 5 concluding with specific recommendations for such future research.
Over the past decades, there has been a growing interest in ‘extreme events’ owing to the increasing threats that climate-related extremes such as floods, heatwaves, droughts, etc., pose to society. While extreme events have diverse definitions across various disciplines, ranging from earth science to neuroscience, they are characterized mainly as dynamic occurrences within a limited time frame that impedes the normal functioning of a system. Although extreme events are rare in occurrence, it has been found in various hydro-meteorological and physiological time series (e.g., river flows, temperatures, heartbeat intervals) that they may exhibit recurrent behavior, i.e., do not end the lifetime of the system. The aim of this thesis to develop some
sophisticated methods to study various properties of extreme events.
One of the main challenges in analyzing such extreme event-like time series is that they have large temporal gaps due to the paucity of the number of observations of extreme events. As a result, existing time series analysis tools are usually not helpful to decode the underlying
information. I use the edit distance (ED) method to analyze extreme event-like time series in their unaltered form. ED is a specific distance metric, mainly designed to measure the similarity/dissimilarity between point process-like data. I combine ED with recurrence plot techniques to identify the recurrence property of flood events in the Mississippi River in the United States. I also use recurrence quantification analysis to show the deterministic properties
and serial dependency in flood events.
After that, I use this non-linear similarity measure (ED) to compute the pairwise dependency in extreme precipitation event series. I incorporate the similarity measure within the framework of complex network theory to study the collective behavior of climate extremes. Under this architecture, the nodes are defined by the spatial grid points of the given spatio-temporal climate dataset. Each node is associated with a time series corresponding to the temporal evolution
of the climate observation at that grid point. Finally, the network links are functions of the pairwise statistical interdependence between the nodes. Various network measures, such as degree, betweenness centrality, clustering coefficient, etc., can be used to quantify the network’s topology. We apply the methodology mentioned above to study the spatio-temporal coherence pattern of extreme rainfall events in the United States and the Ganga River basin, which reveals its relation to various climate processes and the orography of the region.
The identification of precursors associated with the occurrence of extreme events in the near future is extremely important to prepare the masses for an upcoming disaster and mitigate the potential risks associated with such events. Under this motivation, I propose an in-data prediction recipe for predicting the data structures that typically occur prior to extreme events using the Echo state network, a type of Recurrent Neural Network which is a part of the reservoir
computing framework. However, unlike previous works that identify precursory structures in the same variable in which extreme events are manifested (active variable), I try to predict these structures by using data from another dynamic variable (passive variable) which does not show large excursions from the nominal condition but carries imprints of these extreme events. Furthermore, my results demonstrate that the quality of prediction depends on the magnitude
of events, i.e., the higher the magnitude of the extreme, the better is its predictability skill. I show quantitatively that this is because the input signals collectively form a more coherent pattern for an extreme event of higher magnitude, which enhances the efficiency of the machine to predict the forthcoming extreme events.
Text is a ubiquitous entity in our world and daily life. We encounter it nearly everywhere in shops, on the street, or in our flats. Nowadays, more and more text is contained in digital images. These images are either taken using cameras, e.g., smartphone cameras, or taken using scanning devices such as document scanners. The sheer amount of available data, e.g., millions of images taken by Google Streetview, prohibits manual analysis and metadata extraction. Although much progress was made in the area of optical character recognition (OCR) for printed text in documents, broad areas of OCR are still not fully explored and hold many research challenges. With the mainstream usage of machine learning and especially deep learning, one of the most pressing problems is the availability and acquisition of annotated ground truth for the training of machine learning models because obtaining annotated training data using manual annotation mechanisms is time-consuming and costly. In this thesis, we address of how we can reduce the costs of acquiring ground truth annotations for the application of state-of-the-art machine learning methods to optical character recognition pipelines. To this end, we investigate how we can reduce the annotation cost by using only a fraction of the typically required ground truth annotations, e.g., for scene text recognition systems. We also investigate how we can use synthetic data to reduce the need of manual annotation work, e.g., in the area of document analysis for archival material. In the area of scene text recognition, we have developed a novel end-to-end scene text recognition system that can be trained using inexact supervision and shows competitive/state-of-the-art performance on standard benchmark datasets for scene text recognition. Our method consists of two independent neural networks, combined using spatial transformer networks. Both networks learn together to perform text localization and text recognition at the same time while only using annotations for the recognition task. We apply our model to end-to-end scene text recognition (meaning localization and recognition of words) and pure scene text recognition without any changes in the network architecture.
In the second part of this thesis, we introduce novel approaches for using and generating synthetic data to analyze handwriting in archival data. First, we propose a novel preprocessing method to determine whether a given document page contains any handwriting. We propose a novel data synthesis strategy to train a classification model and show that our data synthesis strategy is viable by evaluating the trained model on real images from an archive. Second, we introduce the new analysis task of handwriting classification. Handwriting classification entails classifying a given handwritten word image into classes such as date, word, or number. Such an analysis step allows us to select the best fitting recognition model for subsequent text recognition; it also allows us to reason about the semantic content of a given document page without the need for fine-grained text recognition and further analysis steps, such as Named Entity Recognition. We show that our proposed approaches work well when trained on synthetic data. Further, we propose a flexible metric learning approach to allow zero-shot classification of classes unseen during the network’s training. Last, we propose a novel data synthesis algorithm to train off-the-shelf pixel-wise semantic segmentation networks for documents. Our data synthesis pipeline is based on the famous Style-GAN architecture and can synthesize realistic document images with their corresponding segmentation annotation without the need for any annotated data!
Core-shell upconversion nanoparticles - investigation of dopant intermixing and surface modification
(2022)
Frequency upconversion nanoparticles (UCNPs) are inorganic nanocrystals capable to up-convert incident photons of the near-infrared electromagnetic spectrum (NIR) into higher energy photons. These photons are re-emitted in the range of the visible (Vis) and even ultraviolet (UV) light. The frequency upconversion process (UC) is realized with nanocrystals doped with trivalent lanthanoid ions (Ln(III)). The Ln(III) ions provide the electronic (excited) states forming a ladder-like electronic structure for the Ln(III) electrons in the nanocrystals. The absorption of at least two low energy photons by the nanoparticle and the subsequent energy transfer to one Ln(III) ion leads to the promotion of one Ln(III) electron into higher excited electronic states. One high energy photon will be emitted during the radiative relaxation of the electron in the excited state back into the electronic ground state of the Ln(III) ion. The excited state electron is the result of the previous absorption of at least two low energy photons.
The UC process is very interesting in the biological/medical context. Biological samples (like organic tissue, blood, urine, and stool) absorb high-energy photons (UV and blue light) more strongly than low-energy photons (red and NIR light). Thanks to a naturally occurring optical window, NIR light can penetrate deeper than UV light into biological samples. Hence, UCNPs in bio-samples can be excited by NIR light. This possibility opens a pathway for in vitro as well as in vivo applications, like optical imaging by cell labeling or staining of specific organic tissue. Furthermore, early detection and diagnosis of diseases by predictive and diagnostic biomarkers can be realized with bio-recognition elements being labeled to the UCNPs. Additionally, "theranostic" becomes possible, in which the identification and the treatment of a disease are tackled simultaneously.
For this to succeed, certain parameters for the UCNPs must be met: high upconversion efficiency, high photoluminescence quantum yield, dispersibility, and dispersion stability in aqueous media, as well as availability of functional groups to introduce fast and easy bio-recognition elements. The UCNPs used in this work were prepared with a solvothermal decomposition synthesis yielding in particles with NaYF4 or NaGdF4 as host lattice. They have been doped with the Ln(III) ions Yb3+ and Er3+, which is only one possible upconversion pair. Their upconversion efficiency and photoluminescence quantum yield were improved by adding a passivating shell to reduce surface quenching.
However, the brightness of core-shell UCNPs stays behind the expectations compared to their bulk material (being at least μm-sized particles). The core-shell structures are not clearly separated from each other, which is a topic in literature. Instead, there is a transition layer between the core and the shell structure, which relates to the migration of the dopants within the host lattice during the synthesis. The ion migration has been examined by time-resolved laser spectroscopy and the interlanthanoid resonance energy transfer (LRET) in the two different host lattices from above. The results are
presented in two publications, which dealt with core-shell-shell structured nanoparticles. The core is doped with the LRET-acceptor (either Nd3+ or Pr3+). The intermediate shell serves as an insulation shell of pure host lattice material, whose shell thickness has been varied within one set of samples having the same composition, so that the spatial separation of LRET-acceptor and -donor changes. The outer shell with the same host lattice is doped with the LRET-donor (Eu3+). The effect of the increasing insulation shell thickness is significant, although the LRET cannot be suppressed completely.
Next to the Ln(III) migration within a host lattice, various phase transfer reactions were investigated in order to subsequently perform surface modifications for bioapplications. One result out of this research has been published using a promising ligand, that equips the UCNP with bio-modifiable groups and has good potential for bio-medical applications. This particular ligand mimics natural occurring mechanisms of mussel protein adhesion and of blood coagulation, which is why the UCNPs are encapsulated very effectively. At the same time, bio-functional groups are introduced. In a proof-of-concept, the encapsulated UCNP has been coupled successfully with a dye (which is representative for a biomarker) and the system’s photoluminescence properties have been investigated.
Individuals have an intrinsic need to express themselves to other humans within a given community by sharing their experiences, thoughts, actions, and opinions. As a means, they mostly prefer to use modern online social media platforms such as Twitter, Facebook, personal blogs, and Reddit. Users of these social networks interact by drafting their own statuses updates, publishing photos, and giving likes leaving a considerable amount of data behind them to be analyzed. Researchers recently started exploring the shared social media data to understand online users better and predict their Big five personality traits: agreeableness, conscientiousness, extraversion, neuroticism, and openness to experience. This thesis intends to investigate the possible relationship between users’ Big five personality traits and the published information on their social media profiles. Facebook public data such as linguistic status updates, meta-data of likes objects, profile pictures, emotions, or reactions records were adopted to address the proposed research questions. Several machine learning predictions models were constructed with various experiments to utilize the engineered features correlated with the Big 5 Personality traits. The final predictive performances improved the prediction accuracy compared to state-of-the-art approaches, and the models were evaluated based on established benchmarks in the domain. The research experiments were implemented while ethical and privacy points were concerned. Furthermore, the research aims to raise awareness about privacy between social media users and show what third parties can reveal about users’ private traits from what they share and act on different social networking platforms.
In the second part of the thesis, the variation in personality development is studied within a cross-platform environment such as Facebook and Twitter platforms. The constructed personality profiles in these social platforms are compared to evaluate the effect of the used platforms on one user’s personality development. Likewise, personality continuity and stability analysis are performed using two social media platforms samples. The implemented experiments are based on ten-year longitudinal samples aiming to understand users’ long-term personality development and further unlock the potential of cooperation between psychologists and data scientists.
As of late, epidemiological studies have highlighted a strong association of dairy intake with lower disease risk, and similarly with an increased amount of odd-chain fatty acids (OCFA). While the OCFA also demonstrate inverse associations with disease incidence, the direct dietary sources and mode of action of the OCFA remain poorly understood.
The overall aim of this thesis was to determine the impact of two main fractions of dairy, milk fat and milk protein, on OCFA levels and their influence on health outcomes under high-fat (HF) diet conditions. Both fractions represent viable sources of OCFA, as milk fats contain a significant amount of OCFA and milk proteins are high in branched chain amino acids (BCAA), namely valine (Val) and isoleucine (Ile), which can produce propionyl-CoA (Pr-CoA), a precursor for endogenous OCFA synthesis, while leucine (Leu) does not. Additionally, this project sought to clarify the specific metabolic effects of the OCFA heptadecanoic acid (C17:0).
Both short-term and long-term feeding studies were performed using male C57BL/6JRj mice fed HF diets supplemented with milk fat or C17:0, as well as milk protein or individual BCAA (Val; Leu) to determine their influences on OCFA and metabolic health. Short-term feeding revealed that both milk fractions induce OCFA in vivo, and the increases elicited by milk protein could be, in part, explained by Val intake. In vitro studies using primary hepatocytes further showed an induction of OCFA after Val treatment via de novo lipogenesis and increased α-oxidation. In the long-term studies, both milk fat and milk protein increased hepatic and circulating OCFA levels; however, only milk protein elicited protective effects on adiposity and hepatic fat accumulation—likely mediated by the anti-obesogenic effects of an increased Leu intake. In contrast, Val feeding did not increase OCFA levels nor improve obesity, but rather resulted in glucotoxicity-induced insulin resistance in skeletal muscle mediated by its metabolite 3-hydroxyisobutyrate (3-HIB). Finally, while OCFA levels correlated with improved health outcomes, C17:0 produced negligible effects in preventing HF-diet induced health impairments.
The results presented herein demonstrate that the beneficial health outcomes associated with dairy intake are likely mediated through the effects of milk protein, while OCFA levels are likely a mere association and do not play a significant causal role in metabolic health under HF conditions. Furthermore, the highly divergent metabolic effects of the two BCAA, Leu and Val, unraveled herein highlight the importance of protein quality.
Giros Topográficos
(2022)
Giros topográficos explora las producciones simbólicas del espacio en una serie de textos narrativos publicados desde el cambio de milenio en América Latina. Retomando los planteos teóricos del spatial turn y de la geocrítica, el estudio aborda las topografías literarias desde cuatro ángulos que exceden y transforman los límites territoriales y nacionales: dinámicas de hiperconectividad mediática y movilidad acelerada; genealogías afectivas; ecologías urbanas; y representaciones de la alteridad.
A partir del análisis de obras de Lina Meruane, Guillermo Fadanelli, Andrés Neuman, Andrea Jeftanovic, Sergio Chejfech y Bernardo Carvalho, entre otros, el libro señala los flujos, ambigüedades y tensiones proyectadas por las nuevas comunidades imaginadas del s.XXI. Con ello, el ensayo busca ofrecer un aporte para repensar el estatus de la literatura latinoamericana en el marco de su globalización avanzada y la consecuente consolidación de espacios de enunciación translocalizados.
The negative impact of crude oil on the environment has led to a necessary transition toward alternative, renewable, and sustainable resources. In this regard, lignocellulosic biomass (LCB) is a promising renewable and sustainable alternative to crude oil for the production of fine chemicals and fuels in a so-called biorefinery process. LCB is composed of polysaccharides (cellulose and hemicellulose), as well as aromatics (lignin). The development of a sustainable and economically advantageous biorefinery depends on the complete and efficient valorization of all components. Therefore, in the new generation of biorefinery, the so-called biorefinery of type III, the LCB feedstocks are selectively deconstructed and catalytically transformed into platform chemicals. For this purpose, the development of highly stable and efficient catalysts is crucial for progress toward viability in biorefinery. Furthermore, a modern and integrated biorefinery relies on process and reactor design, toward more efficient and cost-effective methodologies that minimize waste. In this context, the usage of continuous flow systems has the potential to provide safe, sustainable, and innovative transformations with simple process integration and scalability for biorefinery schemes.
This thesis addresses three main challenges for future biorefinery: catalyst synthesis, waste feedstock valorization, and usage of continuous flow technology. Firstly, a cheap, scalable, and sustainable approach is presented for the synthesis of an efficient and stable 35 wt.-% Ni catalyst on highly porous nitrogen-doped carbon support (35Ni/NDC) in pellet shape. Initially, the performance of this catalyst was evaluated for the aqueous phase hydrogenation of LCB-derived compounds such as glucose, xylose, and vanillin in continuous flow systems. The 35Ni/NDC catalyst exhibited high catalytic performances in three tested hydrogenation reactions, i.e., sorbitol, xylitol, and 2-methoxy-4-methylphenol with yields of 82 mol%, 62 mol%, and 100 mol% respectively. In addition, the 35Ni/NDC catalyst exhibited remarkable stability over a long time on stream in continuous flow (40 h). Furthermore, the 35Ni/NDC catalyst was combined with commercially available Beta zeolite in a dual–column integrated process for isosorbide production from glucose (yield 83 mol%).
Finally, 35Ni/NDC was applied for the valorization of industrial waste products, namely sodium lignosulfonate (LS) and beech wood sawdust (BWS) in continuous flow systems. The LS depolymerization was conducted combining solvothermal fragmentation of water/alcohol mixtures (i.e.,methanol/water and ethanol/water) with catalytic hydrogenolysis/hydrogenation (SHF). The depolymerization was found to occur thermally in absence of catalyst with a tunable molecular weight according to temperature. Furthermore, the SHF generated an optimized cumulative yield of lignin-derived phenolic monomers of 42 mg gLS-1. Similarly, a solvothermal and reductive catalytic fragmentation (SF-RCF) of BWS was conducted using MeOH and MeTHF as a solvent. In this case, the optimized total lignin-derived phenolic monomers yield was found of 247 mg gKL-1.
Sustainable urban growth
(2022)
This dissertation explores the determinants for sustainable and socially optimalgrowth in a city. Two general equilibrium models establish the base for this evaluation, each adding its puzzle piece to the urban sustainability discourse and examining the role of non-market-based and market-based policies for balanced growth and welfare improvements in different theory settings. Sustainable urban growth either calls for policy actions or a green energy transition. Further, R&D market failures can pose severe challenges to the sustainability of urban growth and the social optimality of decentralized allocation decisions. Still, a careful (holistic) combination of policy instruments can achieve sustainable growth and even be first best.
Technological progress allows for producing ever more complex predictive models on the basis of increasingly big datasets. For risk management of natural hazards, a multitude of models is needed as basis for decision-making, e.g. in the evaluation of observational data, for the prediction of hazard scenarios, or for statistical estimates of expected damage. The question arises, how modern modelling approaches like machine learning or data-mining can be meaningfully deployed in this thematic field. In addition, with respect to data availability and accessibility, the trend is towards open data. Topic of this thesis is therefore to investigate the possibilities and limitations of machine learning and open geospatial data in the field of flood risk modelling in the broad sense. As this overarching topic is broad in scope, individual relevant aspects are identified and inspected in detail.
A prominent data source in the flood context is satellite-based mapping of inundated areas, for example made openly available by the Copernicus service of the European Union. Great expectations are directed towards these products in scientific literature, both for acute support of relief forces during emergency response action, and for modelling via hydrodynamic models or for damage estimation. Therefore, a focus of this work was set on evaluating these flood masks. From the observation that the quality of these products is insufficient in forested and built-up areas, a procedure for subsequent improvement via machine learning was developed. This procedure is based on a classification algorithm that only requires training data from a particular class to be predicted, in this specific case data of flooded areas, but not of the negative class (dry areas). The application for hurricane Harvey in Houston shows the high potential of this method, which depends on the quality of the initial flood mask.
Next, it is investigated how much the predicted statistical risk from a process-based model chain is dependent on implemented physical process details. Thereby it is demonstrated what a risk study based on established models can deliver. Even for fluvial flooding, such model chains are already quite complex, though, and are hardly available for compound or cascading events comprising torrential rainfall, flash floods, and other processes. In the fourth chapter of this thesis it is therefore tested whether machine learning based on comprehensive damage data can offer a more direct path towards damage modelling, that avoids explicit conception of such a model chain. For that purpose, a state-collected dataset of damaged buildings from the severe El Niño event 2017 in Peru is used. In this context, the possibilities of data-mining for extracting process knowledge are explored as well. It can be shown that various openly available geodata sources contain useful information for flood hazard and damage modelling for complex events, e.g. satellite-based rainfall measurements, topographic and hydrographic information, mapped settlement areas, as well as indicators from spectral data. Further, insights on damaging processes are discovered, which mainly are in line with prior expectations. The maximum intensity of rainfall, for example, acts stronger in cities and steep canyons, while the sum of rain was found more informative in low-lying river catchments and forested areas. Rural areas of Peru exhibited higher vulnerability in the presented study compared to urban areas. However, the general limitations of the methods and the dependence on specific datasets and algorithms also become obvious.
In the overarching discussion, the different methods – process-based modelling, predictive machine learning, and data-mining – are evaluated with respect to the overall research questions. In the case of hazard observation it seems that a focus on novel algorithms makes sense for future research. In the subtopic of hazard modelling, especially for river floods, the improvement of physical models and the integration of process-based and statistical procedures is suggested. For damage modelling the large and representative datasets necessary for the broad application of machine learning are still lacking. Therefore, the improvement of the data basis in the field of damage is currently regarded as more important than the selection of algorithms.