Recent Submissions
Cross-lingual word embeddings with multi-sense representations
(2024) Shim, Soh-Eun
Cross-lingual word embeddings have been found to be useful in aiding cross-lingual transfer, but work in this line of research has to date rarely addressed the monosemy constraint of static word embeddings in depth, where the collapse of multiple meanings into one form might arguably lead to subpar alignments. In this thesis, we address this gap by examining potential approaches towards the incorporation of sense information into cross-lingual alignment. We explore in specfic two variants of cross-lingual multi-sense alignment: one in which we employ the method of embedding the senses of each word as a Gaussian mixture (Athiwaratkun and Wilson, 2017), where the assumption is that multi-sense embeddings as a basis for alignment may help mitigate the meaning conflation deficiency (Camacho-Collados and Pilehvar, 2018), and in turn help improve isomorphism between vector spaces (Ruder et al., 2019). Our second method explores learning a cross-lingual multi-sense embedding space by reversing the order: we cross-lingually align uni-sense word embeddings, and attempt multi-sense enrichment as a postprocessing step by retrofitting (Pilehvar and Collier, 2016) the embedding on the Open Multilingual Wordnet (Bond et al., 2023). We observe that our model is capable of fine-grained cross-lingual semantic distinctions, where our model successfully identifies colexifications without cross-lingual supervision.
Improving the generalisability of fake audio detection
(2024) Lavrynovska, Viktoria
The rapid advancements in neural speech synthesis have enabled the generation of deepfake audios that are increasingly indistinguishable from real voice recordings. Automatic fake audio detection is an emerging research area aiming to develop a reliable means of distinguishing between real and synthetic speech. Lacking generalisability of the detection models on unseen data is an issue that is commonly observed. One of the contributions of this work is the investigation of generalisability across different languages. MesoInception-4 trained on the ASVspoof19 anti-spoofing dataset builds the foundation of our detection model. The use of Mel-Frequency Cepstral Coefficients has been found to be superior to Whisper features for the cross-lingual task. Notably, our model exhibits robust performance on all evaluated languages, despite being trained exclusively on English data, and shows no evidence of language dependency or correlation with the speech quality of the language subsets. However, the findings reveal that the detection model fails to generalise well on the In-the-Wild dataset. We identify that reducing the length of audio clips and fine-tuning specific inception modules can alleviate these issues to some degree. Conversely, augmenting training data with various real-world noises from the MUSAN corpus did not significantly enhance generalisability, and the inclusion of pink noise and silence led to performance degradation on In-the-Wild data. In summary, the findings highlight the complexity of fake audio detection and underscore the importance of further research to elucidate the factors influencing performance and generalisability of detection systems.
An attribution method for classification tasks in Siamese models
(2024) Liu, Mindong
Explaining the contribution of tokens on classification results in the classification task of two sentences is a challenging problem in natural language processing (NLP). This thesis studies the use of the Integrated Jacobians (IJ) in interpreting multi-class classification models with Siamese models, particularly its application in Natural Language Inference (NLI). The NLI task requires models to understand the logical relationships between two sentences, posing challenges for model interpretability. To address the fact that the original Siamese model was primarily designed for regression tasks, the thesis first expanded Siamese models for classification tasks with bilinear similarity while ensuring that the IJ methods can be utilized. It then adapts two forms of the IJ methods: exact IJ and approximate IJ, to work with newly extended Siamese models.
To validate the effectiveness of the extended Siamese models using the IJ meth ods, the thesis conducted experiments on the AllNLI dataset under sentence-BERT framework. The thesis employed four different model configurations and applied both IJ methods to these models. The experimental results demonstrate that the IJ methods effectively provide explanations for us.
Finally, the thesis examined the consistency between the explanations provided by the IJ methods and semantic relationships at the lexical and span levels using datasets WordNet and SpanEX. In the analysis, the IJ methods show that the models capture semantic relationships between words and spans, and there is a correlation between these relationships and the model’s predictions. This finding supports the use of the IJ methods to explain the decisions of NLP models.
Applying models of neighborhood densities to predict multiword expression meaning change
(2025) Lazarus, Len-Noah
Polysemie - die Eigenschaft eines Wortes, mehrere Bedeutungen zu haben - wandelt sich häufig im Laufe der Zeit, wenn sich der Sprachgebrauch ändert. Zum Beispiel die englische Phrase 'gold mine', die historisch für tatsächliche Minen verwendet wurde, in denen Gold abgebaut wird, bekam später eine zusätzliche Bedeutung, nämlich als Beschreibung von generell wertvollen Dingen. Die Nachbarschaftsdichte ist ein skalarer Wert, der angibt, wie nah die Umgebungsvektoren eines Zielwortvektors in einem konstruierten Vektorraum an diesem liegen. In dieser Arbeit wird die Nachbarschaftsdichte verwendet, um die Veränderung der Anzahl der Wortbedeutungen von polymorphen Worten über Zeit vorherzusagen. Dazu werden Co-Occurrence und Word2Vec Embeddings mit über 3000 Hyperparameterkonfigurationen getestet, um festzustellen, wie diese die Stabilität und Quantität der Vorhersagen beeinflussen. Zur Evaluation der Methode werden zwei große diachrone Korpora in deutscher und englischer Sprache, sowie ein Goldstandard für Nomenkomposita in diesen Korpora verwendet. Die beobachteten Auswertungen deuten auf den Erfolg der Verwendung von Nachbarschaftsdichte hin, mit der interessanten Beobachtung, dass Co-Occurrence Vektoren in der Lage sind, die Änderung in der Anzahl der Bedeutungen über Zeit vorherzusagen, aber unfähig zu sein scheinen, die Anzahl der Wortbedeutungen in einer individuellen Zeitepoche anzunähern, während Word2Vec für beides gute Vorhersagen macht.
Decoding the Baroque : development of a novel dataset for transformer-based harmonic analysis of flute music
(2025) Cunningham, Andrea Maria
The computational modeling of functional harmony in the field of Music Information Retrieval (MIR) faces significant challenges when addressing the intricate structural relationships inherent in historical styles like Baroque music. This research introduces a novel dataset of Baroque flute sonatas with detailed functional harmony annotations spanning church and chamber traditions. A specialized Transformer-based model for automatic melody harmonization is developed and evaluated on its ability to generate stylistically appropriate harmonizations across composers and forms. The study proposes that a segmentation-aware Transformer model with innovative attention mechanisms can produce stylistically coherent harmonizations while revealing tensions between period conventions and composer-specific idioms. By examining Baroque flute instrumental music beyond Bach chorales, this study addresses a critical gap in computational musicology and demonstrates how sequence-to-sequence architectures capture hierarchical structures analogous to linguistics. This thesis describes the data curation process of the Baroque flute sonata dataset, presents the enhanced Transformer model, outlines the methodology, and analyzes results concerning harmonic accuracy, stylistic coherence, and musicological implications.
The role of spacer length in macrocyclization reactions under confinement
(2024) Nandeshwar, Muneshwar; Weisser, Kilian; Ziegler, Felix; Frey, Wolfgang; Buchmeiser, Michael R.
We studied the influence of the distance of olefin metathesis catalysts from the inner surface of a mesoporous support on macrocyclization and Z‐selectivity under confinement. For these purposes, the cationic molybdenum imido alkylidene N‐heterocyclic carbene (NHC) catalysts [Mo(N‐(2‐tBu‐C6H4)(1‐mesityl‐3‐(3‐trimethoxysilylprop‐1‐yl)‐imidazol‐2‐ylidene)(CHCMe2Ph)(MeCN)Br+ B(ArF)4-] Mo2, [Mo(N‐(2‐tBu‐C6H4)(1‐mesityl‐3‐(3‐trimethoxysilylprop‐1‐yl)‐imidazol‐2‐ylidene)(CHCMe2Ph)(MeCN)OTf+ B(ArF)4-] Mo3, [Mo(N‐(2,6‐Me2‐C6H3)(1‐mesityl‐3‐(3‐trimethoxysilylprop‐1‐yl)‐imidazol‐2‐ylidene)(CHCMe2Ph)(MeCN)Br+ B(ArF)4-] Mo5, and [Mo(N‐(2,6‐iPr2‐C6H3)(1‐mesityl‐3‐(3‐trimethoxysilylprop‐1‐yl)‐imidazol‐2‐ylidene)(CHCMe2Ph)(MeCN)+Br B(ArF)4-] Mo7 (B(ArF)4 = tetrakis[3,5‐bis(trifluoromethyl)phenyl]borate), all containing a trimethoxysilylpropyl tether, were selectively immobilized inside the mesopores of SBA‐15. Under confinement, both macro(mono)cyclization (MMC) and Z‐selectivity were higher than in solution but lower than with catalysts directly bound to the surface of the mesoporous supports. These findings are in agreement with existing theoretical models on substrate and product distribution in mesopores, which suggest that the highest substrate concentration is found at the pore wall and that it increases with decreasing pore diameter.
Rayleigh invariance allows the estimation of effective CO2 fluxes due to convective dissolution into water‐filled fractures
(2025) Keim, Leon; Class, Holger
Convective dissolution of CO2 is a well‐known mechanism in geological storage of CO2. It is triggered by gravitational instability which leads to the onset of free convection. The phenomenon is well studied in porous media, such as saline aquifers, and the literature provides substantial evidence that onset times and effective flux rates can be estimated based on a characterization of instabilities that uses the Darcy velocity. This work extends the study of convective dissolution to open water‐filled fractures, where non‐Darcy regimes govern the induced flow processes. Numerical simulations using a Navier‐Stokes model with fluid density dependent on dissolved CO2 concentration were used to compute scenario‐specific results for effective CO2 entry rates into an idealized fracture with varying aperture, temperature, and CO2 concentration at the gas‐water interface. The results were analyzed in terms of dimensionless quantities. They revealed a Rayleigh invariance of the effective CO2 flux after the complete formation of a quasi‐stationary velocity profile, that is, after a certain entry length. Hence, this invariance can be exploited to estimate the effective CO2 entry rates, which can then be used, in perspective, in upscaled models. We have studied convective CO2 dissolution for two different fracture settings; the first one relates to karstification scenarios, where CO2 is the dominant driving force, and were stagnant‐water conditions in fractures have not yet received attention to date. The second setting is inspired from geological CO2 storage, where the literature provides only studies on convective CO2 dissolution for porous‐media flow with Darcy regimes.
Enhanced oxidation resistance of Pt‐containing Inconel 718 alloy through facilitated formation of protective chromia
(2024) Zheng, Jianshu; Zhang, Boning; Wang, Guowei; Liu, Lan; Shen, Chao; Song, Zhuorui; Zheng, Lei
This study investigates the influence of Pt addition on the oxidation behavior of a Cr2O3‐forming superalloy. Inconel 718 (IN718) alloys with varying Pt content were prepared and subjected to isothermal oxidation tests. The results demonstrate that Pt significantly enhances the oxidation resistance of IN718, as evidenced by reduced weight gain, thinner oxide layers, and smaller oxide particles. Pt addition also increases the activation energy for both initial interface oxidation and ion diffusion during long‐term oxidation. Furthermore, Pt promotes the formation of a Cr2O3 layer while suppressing the formation of other undesirable oxides, resulting in a more cohesive and stable oxide layer. The improved oxidation resistance is attributed to two key factors: during the initial oxidation stage, Pt, as a noble element, reduces the activity of the primary oxide‐forming element Cr to oxidative environments, thereby lowering its susceptibility to initial oxidation at the metal-oxidant interface. During long‐term oxidation, Pt preferentially substitutes for Ni in major phases such as γ‐Ni(Cr,Fe) and γ′‐Ni3(Al,Ti), locally increasing the Cr composition. This promotes Cr oxidation, effectively suppressing the oxidation of Ni or Fe. These findings suggest that Pt addition is a promising approach for enhancing oxidation resistance in alloy design.
Dissertationen und Habilitationsschriften der Universität Stuttgart 2021/2, Juli - Dezember 2021
(Stuttgart : Universitätsbibliothek, 2022)
Dissertationen und Habilitationsschriften der Universität Stuttgart 2021/1, Januar - Juni 2021
(Stuttgart : Universitätsbibliothek, 2021)