Mathias Müller

Zifan Jiang, Gerard Sant, Amit Moryossef et al.

We present SignCLIP, which re-purposes CLIP (Contrastive Language-Image Pretraining) to project spoken language text and sign language videos, two classes of natural languages of distinct modalities, into the same space. SignCLIP is an efficient method of learning useful visual representations for sign language processing from large-scale, multilingual video-text pairs, without directly optimizing for a specific task or sign language which is often of limited size. We pretrain SignCLIP on Spreadthesign, a prominent sign language dictionary consisting of ~500 thousand video clips in up to 44 sign languages, and evaluate it with various downstream datasets. SignCLIP discerns in-domain signing with notable text-to-video/video-to-text retrieval accuracy. It also performs competitively for out-of-domain downstream tasks such as isolated sign language recognition upon essential few-shot prompting or fine-tuning. We analyze the latent space formed by the spoken language text and sign language poses, which provides additional linguistic insights. Our code and models are openly available.

Mathias Müller, Zifan Jiang, Amit Moryossef et al.

Automatic sign language processing is gaining popularity in Natural Language Processing (NLP) research (Yin et al., 2021). In machine translation (MT) in particular, sign language translation based on glosses is a prominent approach. In this paper, we review recent works on neural gloss translation. We find that limitations of glosses in general and limitations of specific datasets are not discussed in a transparent manner and that there is no common standard for evaluation. To address these issues, we put forward concrete recommendations for future research on gloss translation. Our suggestions advocate awareness of the inherent limitations of gloss-based approaches, realistic datasets, stronger baselines and convincing evaluation.

Zifan Jiang, Amit Moryossef, Mathias Müller et al.

This paper presents work on novel machine translation (MT) systems between spoken and signed languages, where signed languages are represented in SignWriting, a sign language writing system. Our work seeks to address the lack of out-of-the-box support for signed languages in current MT systems and is based on the SignBank dataset, which contains pairs of spoken language text and SignWriting content. We introduce novel methods to parse, factorize, decode, and evaluate SignWriting, leveraging ideas from neural factored MT. In a bilingual setup--translating from American Sign Language to (American) English--our method achieves over 30 BLEU, while in two multilingual setups--translating in both directions between spoken languages and signed languages--we achieve over 20 BLEU. We find that common MT techniques used to improve spoken language translation similarly affect the performance of sign language translation. These findings validate our use of an intermediate text representation for signed languages to include them in natural language processing research.

Amit Moryossef, Zifan Jiang, Mathias Müller et al.

Sign language segmentation is a crucial task in sign language processing systems. It enables downstream tasks such as sign recognition, transcription, and machine translation. In this work, we consider two kinds of segmentation: segmentation into individual signs and segmentation into phrases, larger units comprising several signs. We propose a novel approach to jointly model these two tasks. Our method is motivated by linguistic cues observed in sign language corpora. We replace the predominant IO tagging scheme with BIO tagging to account for continuous signing. Given that prosody plays a significant role in phrase boundaries, we explore the use of optical flow features. We also provide an extensive analysis of hand shapes and 3D hand normalization. We find that introducing BIO tagging is necessary to model sign boundaries. Explicitly encoding prosody by optical flow improves segmentation in shallow models, but its contribution is negligible in deeper models. Careful tuning of the decoding algorithm atop the models further improves the segmentation quality. We demonstrate that our final models generalize to out-of-domain video content in a different signed language, even under a zero-shot setting. We observe that including optical flow and 3D hand normalization enhances the robustness of the model in this context.

Amit Moryossef, Mathias Müller, Rebecka Fahrni

Managing and analyzing pose data is a complex task, with challenges ranging from handling diverse file structures and data types to facilitating effective data manipulations such as normalization and augmentation. This paper presents \texttt{pose-format}, a comprehensive toolkit designed to address these challenges by providing a unified, flexible, and easy-to-use interface. The library includes a specialized file format that encapsulates various types of pose data, accommodating multiple individuals and an indefinite number of time frames, thus proving its utility for both image and video data. Furthermore, it offers seamless integration with popular numerical libraries such as NumPy, PyTorch, and TensorFlow, thereby enabling robust machine-learning applications. Through benchmarking, we demonstrate that our \texttt{.pose} file format offers vastly superior performance against prevalent formats like OpenPose, with added advantages like self-contained pose specification. Additionally, the library includes features for data normalization, augmentation, and easy-to-use visualization capabilities, both in Python and Browser environments. \texttt{pose-format} emerges as a one-stop solution, streamlining the complexities of pose data management and analysis.

Eric Egli, Noah Mamié, Eyal Liron Dolev et al.

In this study, we use recent stance detection methods to study the stance (for, against or neutral) of statements in official information booklets for voters. Our main goal is to answer the fundamental question: are topics to be voted on presented in a neutral way? To this end, we first train and compare several models for stance detection on a large dataset about Swiss politics. We find that fine-tuning an M-BERT model leads to the best accuracy. We then use our best model to analyze the stance of utterances extracted from the Swiss federal voting booklet concerning the Swiss popular votes of September 2022, which is the main goal of this project. We evaluated the models in both a multilingual as well as a monolingual context for German, French, and Italian. Our analysis shows that some issues are heavily favored while others are more balanced, and that the results are largely consistent across languages. Our findings have implications for the editorial process of future voting booklets and the design of better automated systems for analyzing political discourse. The data and code accompanying this paper are available at https://github.com/ZurichNLP/voting-booklet-bias.

Shester Gueuwou, Sophie Siake, Colin Leong et al.

Advancements in sign language processing have been hindered by a lack of sufficient data, impeding progress in recognition, translation, and production tasks. The absence of comprehensive sign language datasets across the world's sign languages has widened the gap in this field, resulting in a few sign languages being studied more than others, making this research area extremely skewed mostly towards sign languages from high-income countries. In this work we introduce a new large and highly multilingual dataset for sign language translation: JWSign. The dataset consists of 2,530 hours of Bible translations in 98 sign languages, featuring more than 1,500 individual signers. On this dataset, we report neural machine translation experiments. Apart from bilingual baseline systems, we also train multilingual systems, including some that take into account the typological relatedness of signed or spoken languages. Our experiments highlight that multilingual systems are superior to bilingual baselines, and that in higher-resource scenarios, clustering language pairs that are related improves translation quality.

Pascal Pfeiffer, Dmitry Gordeev, Mathias Müller et al.

We present TabH2O, a foundation model for tabular data that performs classification and regression in a single forward pass via in-context learning. TabH2O builds on the TabICL architecture with several key modifications: (1) unified training, a single model handles both classification and regression via a dual-head architecture, eliminating the need for separate models and reducing total pretraining cost; (2) single-stage pretraining, training stability improvements (bounded scalable softmax, inter-stage normalization, learnable residual scaling, logit soft-capping) eliminate the need for multi-stage curriculum learning, enabling training with full-length sequences from the start; and (3) noise-aware pretraining, synthetic datasets include explicit noise dimensions to teach the model robustness to irrelevant features. We evaluate TabH2O v1 (29.2M parameters) on the TALENT benchmark (300 datasets), where it achieves an average rank of 2.55 out of 6 evaluated methods, outperforming tuned CatBoost (4.07), H2O AutoML (4.18), and LightGBM (5.08), competitive with TabPFN v2.6 (2.74), and behind TabICL v2 (2.12), while placing in the top-3 on 81% of the testing datasets across classification and regression tasks.

Amit Moryossef, Mathias Müller, Anne Göhring et al.

Sign language translation systems are complex and require many components. As a result, it is very hard to compare methods across publications. We present an open-source implementation of a text-to-gloss-to-pose-to-video pipeline approach, demonstrating conversion from German to Swiss German Sign Language, French to French Sign Language of Switzerland, and Italian to Italian Sign Language of Switzerland. We propose three different components for the text-to-gloss translation: a lemmatizer, a rule-based word reordering and dropping component, and a neural machine translation system. Gloss-to-pose conversion occurs using data from a lexicon for three different signed languages, with skeletal poses extracted from videos. To generate a sentence, the text-to-gloss system is first run, and the pose representations of the resulting signs are stitched together.

Biao Zhang, Mathias Müller, Rico Sennrich

Despite recent successes with neural models for sign language translation (SLT), translation quality still lags behind spoken languages because of the data scarcity and modality gap between sign video and text. To address both problems, we investigate strategies for cross-modality representation sharing for SLT. We propose SLTUNET, a simple unified neural model designed to support multiple SLTrelated tasks jointly, such as sign-to-gloss, gloss-to-text and sign-to-text translation. Jointly modeling different tasks endows SLTUNET with the capability to explore the cross-task relatedness that could help narrow the modality gap. In addition, this allows us to leverage the knowledge from external resources, such as abundant parallel data used for spoken-language machine translation (MT). We show in experiments that SLTUNET achieves competitive and even state-of-the-art performance on PHOENIX-2014T and CSL-Daily when augmented with MT data and equipped with a set of optimization techniques. We further use the DGS Corpus for end-to-end SLT for the first time. It covers broader domains with a significantly larger vocabulary, which is more challenging and which we consider to allow for a more realistic assessment of the current state of SLT than the former two. Still, SLTUNET obtains improved results on the DGS Corpus. Code is available at https://github.com/bzhangGo/sltunet.

Catherine O'Brien, Gerard Sant, Mathias Müller et al.

Many sign language translation (SLT) systems operate on pose sequences instead of raw video to reduce input dimensionality, improve portability, and partially anonymize signers. The choice of pose estimator is often treated as an implementation detail, with systems defaulting to widely available tools such as MediaPipe Holistic or OpenPose. We present a systematic comparison of pose estimators for pose-based SLT, covering widely used baselines (MediaPipe Holistic, OpenPose) and newer whole-body/high-capacity models (MMPose WholeBody, OpenPifPaf, AlphaPose, SDPose, Sapiens, SMPLest-X). We quantify downstream impact by training a controlled SLT pipeline on RWTH-PHOENIX-Weather 2014 where only the pose representation varies, evaluating with BLEU and BLEURT. To contextualize translation outcomes, we analyze temporal stability, missing hand keypoints, and robustness to occlusion using higher-resolution videos from the Signsuisse dataset. SDPose and Sapiens achieve the best translation performance (BLEU ~11.5), outperforming the common MediaPipe baseline (BLEU ~10). In occlusion cases, Sapiens is correct in all tested instances (15/15), while OpenPifPaf fails in nearly all (1/15) and also yields the weakest translation scores. Estimators that frequently leave out hand keypoints are associated with lower BLEU/BLEURT. We release code that can be used not only to reproduce our experiments, but also considerably lowers the barrier for other researchers to use alternative pose estimators.

Gerard Sant, Zifan Jiang, Carlos Escolano et al.

In recent years, sign language processing (SLP) has gained importance in the general field of Natural Language Processing. However, compared to research on spoken languages, SLP research is hindered by complex ad-hoc code, inadvertently leading to low reproducibility and unfair comparisons. Existing tools that are built for fast and reproducible experimentation, such as Hugging Face, are not flexible enough to seamlessly integrate sign language experiments. This view is confirmed by a survey we conducted among SLP researchers. To address these challenges, we introduce MultimodalHugs, a framework built on top of Hugging Face that enables more diverse data modalities and tasks, while inheriting the well-known advantages of the Hugging Face ecosystem. Even though sign languages are our primary focus, MultimodalHugs adds a layer of abstraction that makes it more widely applicable to other use cases that do not fit one of the standard templates of Hugging Face. We provide quantitative experiments to illustrate how MultimodalHugs can accommodate diverse modalities such as pose estimation data for sign languages, or pixel data for text characters.

Alessandro Mastrototaro, Mathias Müller, Jimmy Olsson

General state-space models (SSMs) are widely used in statistical machine learning and are among the most classical generative models for sequential time-series data. SSMs, comprising latent Markovian states, can be subjected to variational inference (VI), but standard VI methods like the importance-weighted autoencoder (IWAE) lack functionality for streaming data. To enable online VI in SSMs when the observations are received in real time, we propose maximising an IWAE-type variational lower bound on the asymptotic contrast function, rather than the standard IWAE ELBO, using stochastic approximation. Unlike the recursive maximum likelihood method, which directly maximises the asymptotic contrast, our approach, called online sequential IWAE (OSIWAE), allows for online learning of both model parameters and a Markovian recognition model for inferring latent states. By approximating filter state posteriors and their derivatives using sequential Monte Carlo (SMC) methods, we create a particle-based framework for online VI in SSMs. This approach is more theoretically well-founded than recently proposed online variational SMC methods. We provide rigorous theoretical results on the learning objective and a numerical study demonstrating the method's efficiency in learning model parameters and particle proposal kernels.

Mathias Müller, Rico Sennrich

Neural Machine Translation (NMT) currently exhibits biases such as producing translations that are too short and overgenerating frequent words, and shows poor robustness to copy noise in training data or domain shift. Recent work has tied these shortcomings to beam search -- the de facto standard inference algorithm in NMT -- and Eikema & Aziz (2020) propose to use Minimum Bayes Risk (MBR) decoding on unbiased samples instead. In this paper, we empirically investigate the properties of MBR decoding on a number of previously reported biases and failure cases of beam search. We find that MBR still exhibits a length and token frequency bias, owing to the MT metrics used as utility functions, but that MBR also increases robustness against copy noise in the training data and domain shift.

Amit Moryossef, Ioannis Tsochantaridis, Joe Dinn et al.

Signed languages are visual languages produced by the movement of the hands, face, and body. In this paper, we evaluate representations based on skeleton poses, as these are explainable, person-independent, privacy-preserving, low-dimensional representations. Basically, skeletal representations generalize over an individual's appearance and background, allowing us to focus on the recognition of motion. But how much information is lost by the skeletal representation? We perform two independent studies using two state-of-the-art pose estimation systems. We analyze the applicability of the pose estimation systems to sign language recognition by evaluating the failure cases of the recognition models. Importantly, this allows us to characterize the current limitations of skeletal pose estimation approaches in sign language recognition.

Julia Kreutzer, Isaac Caswell, Lisa Wang et al.

With the success of large-scale pre-training and multilingual modeling in Natural Language Processing (NLP), recent years have seen a proliferation of large, web-mined text datasets covering hundreds of languages. We manually audit the quality of 205 language-specific corpora released with five major public datasets (CCAligned, ParaCrawl, WikiMatrix, OSCAR, mC4). Lower-resource corpora have systematic issues: At least 15 corpora have no usable text, and a significant fraction contains less than 50% sentences of acceptable quality. In addition, many are mislabeled or use nonstandard/ambiguous language codes. We demonstrate that these issues are easy to detect even for non-proficient speakers, and supplement the human audit with automatic analyses. Finally, we recommend techniques to evaluate and improve multilingual corpora and discuss potential risks that come with low-quality data releases.

Annette Rios, Mathias Müller, Rico Sennrich

Zero-shot neural machine translation is an attractive goal because of the high cost of obtaining data and building translation systems for new translation directions. However, previous papers have reported mixed success in zero-shot translation. It is hard to predict in which settings it will be effective, and what limits performance compared to a fully supervised system. In this paper, we investigate zero-shot performance of a multilingual EN$\leftrightarrow${FR,CS,DE,FI} system trained on WMT data. We find that zero-shot performance is highly unstable and can vary by more than 6 BLEU between training runs, making it difficult to reliably track improvements. We observe a bias towards copying the source in zero-shot translation, and investigate how the choice of subword segmentation affects this bias. We find that language-specific subword segmentation results in less subword copying at training time, and leads to better zero-shot performance compared to jointly trained segmentation. A recent trend in multilingual models is to not train on parallel data between all language pairs, but have a single bridge language, e.g. English. We find that this negatively affects zero-shot translation and leads to a failure mode where the model ignores the language tag and instead produces English output in zero-shot directions. We show that this bias towards English can be effectively reduced with even a small amount of parallel data in some of the non-English pairs.

Franziska Hannß, Esther Lapczyna, Mathias Müller et al.

This case study describes an approach to designing interactive museum installations as a student project with the aim of presenting the research results of the restoration process of paintings to a wide range of visitors. During one and a half years, the Chair of Media Design created five interactive media stations in two lectures to enrich the special exhibition "Veronese: The Cuccina Cycle. The Restored Masterpiece". The project was realised in close communication with the conservators of the Dresden State Art Collections and the employees of the Science and Archaeometric Laboratory of the Dresden University of Fine Arts. The students had to learn about the foreign content and how to translate it into a media-related environment. With suitable teaching methods, we pushed the students towards a deeper understanding of the matter.

Dmitry Gordeev, Philipp Singer, Marios Michailidis et al.

The advent of the COVID-19 pandemic has instigated unprecedented changes in many countries around the globe, putting a significant burden on the health sectors, affecting the macro economic conditions, and altering social interactions amongst the population. In response, the academic community has produced multiple forecasting models, approaches and algorithms to best predict the different indicators of COVID-19, such as the number of confirmed infected cases. Yet, researchers had little to no historical information about the pandemic at their disposal in order to inform their forecasting methods. Our work studies the predictive performance of models at various stages of the pandemic to better understand their fundamental uncertainty and the impact of data availability on such forecasts. We use historical data of COVID-19 infections from 253 regions from the period of 22nd January 2020 until 22nd June 2020 to predict, through a rolling window backtesting framework, the cumulative number of infected cases for the next 7 and 28 days. We implement three simple models to track the root mean squared logarithmic error in this 6-month span, a baseline model that always predicts the last known value of the cumulative confirmed cases, a power growth model and an epidemiological model called SEIRD. Prediction errors are substantially higher in early stages of the pandemic, resulting from limited data. Throughout the course of the pandemic, errors regress slowly, but steadily. The more confirmed cases a country exhibits at any point in time, the lower the error in forecasting future confirmed cases. We emphasize the significance of having a rigorous backtesting framework to accurately assess the predictive power of such models at any point in time during the outbreak which in turn can be used to assign the right level of certainty to these forecasts and facilitate better planning.

Mathias Müller, Annette Rios, Rico Sennrich

Translating text that diverges from the training domain is a key challenge for machine translation. Domain robustness---the generalization of models to unseen test domains---is low for both statistical (SMT) and neural machine translation (NMT). In this paper, we study the performance of SMT and NMT models on out-of-domain test sets. We find that in unknown domains, SMT and NMT suffer from very different problems: SMT systems are mostly adequate but not fluent, while NMT systems are mostly fluent, but not adequate. For NMT, we identify such hallucinations (translations that are fluent but unrelated to the source) as a key reason for low domain robustness. To mitigate this problem, we empirically compare methods that are reported to improve adequacy or in-domain robustness in terms of their effectiveness at improving domain robustness. In experiments on German to English OPUS data, and German to Romansh (a low-resource setting) we find that several methods improve domain robustness. While those methods do lead to higher BLEU scores overall, they only slightly increase the adequacy of translations compared to SMT.

Mathias Müller, Annette Rios, Elena Voita et al.

The translation of pronouns presents a special challenge to machine translation to this day, since it often requires context outside the current sentence. Recent work on models that have access to information across sentence boundaries has seen only moderate improvements in terms of automatic evaluation metrics such as BLEU. However, metrics that quantify the overall translation quality are ill-equipped to measure gains from additional context. We argue that a different kind of evaluation is needed to assess how well models translate inter-sentential phenomena such as pronouns. This paper therefore presents a test suite of contrastive translations focused specifically on the translation of pronouns. Furthermore, we perform experiments with several context-aware models. We show that, while gains in BLEU are moderate for those systems, they outperform baselines by a large margin in terms of accuracy on our contrastive test set. Our experiments also show the effectiveness of parameter tying for multi-encoder architectures.

Gongbo Tang, Mathias Müller, Annette Rios et al.

Recently, non-recurrent architectures (convolutional, self-attentional) have outperformed RNNs in neural machine translation. CNNs and self-attentional networks can connect distant words via shorter network paths than RNNs, and it has been speculated that this improves their ability to model long-range dependencies. However, this theoretical argument has not been tested empirically, nor have alternative explanations for their strong performance been explored in-depth. We hypothesize that the strong performance of CNNs and self-attentional networks could also be due to their ability to extract semantic features from the source text, and we evaluate RNNs, CNNs and self-attention networks on two tasks: subject-verb agreement (where capturing long-range dependencies is required) and word sense disambiguation (where semantic feature extraction is required). Our experimental results show that: 1) self-attentional networks and CNNs do not outperform RNNs in modeling subject-verb agreement over long distances; 2) self-attentional networks perform distinctly better than RNNs and CNNs on word sense disambiguation.