Machine learning (ML) models are increasingly used to support clinical decision-making. However, real-world medical datasets are often noisy, incomplete, and imbalanced, leading to performance disparities across patient subgroups. These differences raise fairness concerns, particularly when they reinforce existing disadvantages for marginalized groups. In this work, we analyze several medical prediction tasks and demonstrate how model performance varies with patient characteristics. While ML models may demonstrate good overall performance, we argue that subgroup-level evaluation is essential before integrating them into clinical workflows. By conducting a performance analysis at the subgroup level, differences can be clearly identified—allowing, on the one hand, for performance disparities to be considered in clinical practice, and on the other hand, for these insights to inform the responsible development of more effective models. Thereby, our work contributes to a practical discussion around the subgroup-sensitive development and deployment of medical ML models and the interconnectedness of fairness and transparency.

In electronic health records, text data is considered a valuable resource as it complements a medical history and may contain information that cannot be easily included in tables. But why does the inclusion of clinical texts as additional input into multimodal models, not always significantly improve the performance of medical decision-support systems? Explainable AI (XAI) might provide the answer. We examine which information in text and structured data influences the performance of models in the context of multimodal decision support for biomedical tasks. Using data from an intensive care unit and targeting a mortality prediction task, we compare information that has been considered relevant by XAI methods to the opinion of a physician.

We investigate the impact of LLMs on political discourse with a particular focus on the influence of generated personas on model responses. We find an echo chamber effect from LLM chatbots when provided with German-language biographical information of politicians and voters in German politics, leading to sycophantic responses and the reinforcement of existing political biases. Findings reveal that personas of certain political party, such as those of the ‘Alternative für Deutschland’ party, exert a stronger influence on LLMs, potentially amplifying extremist views. Unlike prior studies, we cannot corroborate a tendency for larger models to exert stronger sycophantic behaviour. We propose that further development should aim at reducing sycophantic behaviour in LLMs across all sizes and diversifying language capabilities in LLMs to enhance inclusivity.

With the rise of Large Generative AI Models (LGAIMs), disinformation online has become more concerning than ever before. Within the super-election year 2024, the influence of mis- and disinformation can severely influence public opinion. To combat the increasing amount of disinformation online, humans need to be supported by AI-based tools to increase the effectiveness of detecting false content. This paper examines the critical intersection of the AI Act with the deployment of LGAIMs for disinformation detection and the implications from research, deployer, and the user’s perspective. The utilization of LGAIMs for disinformation detection falls under the high-risk category defined in the AI Act, leading to several obligations that need to be followed after the enforcement of the AI Act. Among others, the obligations include risk management, transparency, and human oversight which pose the challenge of finding adequate technical interpretations. Furthermore, the paper articulates the necessity for clear guidelines and standards that enable the effective, ethical, and legally compliant use of AI. The paper contributes to the discourse on balancing technological advancement with ethical and legal imperatives, advocating for a collaborative approach to utilizing LGAIMs in safeguarding information integrity and fostering trust in digital ecosystems.

Modern large language models and chatbots based on them show impressive results in text generation and dialog tasks. At the same time, these models are subject to criticism in many aspects, e.g., they can generate hate speech and untrue and biased content. In this work, we show another problematic feature of such chatbots: they are echo chambers in the sense that they tend to agree with the opinions of their users. Social media, such as Facebook, was criticized for a similar problem and called an echo chamber. We experimentally test five LLM-based chatbots, which we feed with opinionated inputs. We annotate the chatbot answers whether they agree or disagree with the input. All chatbots tend to agree. However, the echo chamber effect is not equally strong. We discuss the differences between the chatbots and make the dataset publicly available.

While text-based medical applications have become increasingly prominent, access to clinicaldata remains a major concern. To resolve this issue, further de-identification and anonymization of the data are required. This might, however, alter the contextual information within the clinical texts and therefore influence the learning and performance of possible language models. This paper systematically analyses the potential effects of various anonymization techniques on the performance of state-of-the-art machine learning models based on several datasets corresponding to five different NLP tasks. On this basis, we derive insightful findings and recommendations concerning text anonymization with regard to the performance of machine learning models. In addition, we present a simple re-identification attack applied to the anonymized text data, which can break the anonymization.

This paper presents a fine-grained test suite for the language pair German–English. The test suite is based on a number of linguistically motivated categories and phenomena and the semi-automatic evaluation is carried out with regular expressions. We describe the creation and implementation of the test suite in detail, providing a full list of all categories and phenomena. Furthermore, we present various exemplary applications of our test suite that have been implemented in the past years, like contributions to the Conference of Machine Translation, the usage of the test suite and MT outputs for quality estimation, and the expansion of the test suite to the language pair Portuguese–English. We describe how we tracked the development of the performance of various systems MT systems over the years with the help of the test suite and which categories and phenomena are prone to resulting in MT errors. For the first time, we also make a large part of our test suite publicly available to the research community.

In recent years, machine learning for clinical decision support has gained more and more attention. In order to introduce such applications into clinical practice, a good performance might be essential, however, the aspect of trust should not be underestimated. For the treating physician using such a system and being (legally) responsible for the decision made, it is particularly important to understand the system’s recommendation. To provide insights into a model’s decision, various techniques from the field of explainability (XAI) have been proposed whose output is often enough not targeted to the domain experts that want to use the model. To close this gap, in this work, we explore how explanations could possibly look like in future. To this end, this work presents a dataset of textual explanations in context of decision support. Within a reader study, human physicians estimated the likelihood of possible negative patient outcomes in the near future and justified each decision with a few sentences. Using those sentences, we created a novel corpus, annotated with different semantic layers. Moreover, we provide an analysis of how those explanations are constructed, and how they change depending on physician, on the estimated risk and also in comparison to an automatic clinical decision support system with feature importance.

This paper describes a test suite submission providing detailed statistics of linguistic performance for the state-of-the-art German-English systems of the Fifth Conference of Machine Translation (WMT20). The analysis covers 107 phenomena organized in 14 categories based on about 5,500 test items, including a manual annotation effort of 45 person hours. Two systems (Tohoku and Huoshan) appear to have significantly better test suite accuracy than the others, although the best system of WMT20 is not significantly better than the one from WMT19 in a macro-average. Additionally, we identify some linguistic phenomena where all systems suffer (such as idioms, resultative predicates and pluperfect), but we are also able to identify particular weaknesses for individual systems (such as quotation marks, lexical ambiguity and sluicing). Most of the systems of WMT19 which submitted new versions this year show improvements.

We present an analysis of 16 state-of-the-art MT systems on German-English based on a linguistically-motivated test suite. The test suite has been devised manually by a team of language professionals in order to cover a broad variety of linguistic phenomena that MT often fails to translate properly. It contains 5,000 test sentences covering 106 linguistic phenomena in 14 categories, with an increased focus on verb tenses, aspects and moods. The MT outputs are evaluated in a semi-automatic way through regular expressions that focus only on the part of the sentence that is relevant to each phenomenon. Through our analysis, we are able to compare systems based on their performance on these categories. Additionally, we reveal strengths and weaknesses of particular systems and we identify grammatical phenomena where the overall performance of MT is relatively low.

The Conference on Computational Natural Language Learning (CoNLL) features a shared task, in which participants train and test their learning systems on the same data sets. In 2017, the task was devoted to learning dependency parsers for a large number of languages, in a real-world setting without any gold-standard annotation on input. All test sets followed a unified annotation scheme, namely that of Universal Dependencies. In this paper, we define the task and evaluation methodology, describe how the data sets were prepared, report and analyze the main results, and provide a brief categorization of the different approaches of the participating systems.

In this document we report on a user-scenario-based evaluation aiming at assessing the performance of machine translation (MT) systems in a real context of use. We describe a sequel of experiments that has been performed to estimate the usefulness of MT and to test if improvements of MT technology lead to better performance in the usage scenario. One goal is to find the best methodology for evaluating the eventual benefit of a machine translation system in an application. The evaluation is based on the QTLeap corpus, a novel multilingual language resource that was collected through a real-life support service via chat. It is composed of naturally occurring utterances produced by users while interacting with a human technician providing answers. The corpus is available in eight different languages: Basque, Bulgarian, Czech, Dutch, English, German, Portuguese and Spanish.

This work addresses the need to aid Machine Translation (MT) development cycles with a complete workflow of MT evaluation methods. Our aim is to assess, compare and improve MT system variants. We hereby report on novel tools and practices that support various measures, developed in order to support a principled and informed approach of MT development. Our toolkit for automatic evaluation showcases quick and detailed comparison of MT system variants through automatic metrics and n-gram feedback, along with manual evaluation via edit-distance, error annotation and task-based feedback.

Human translators are the key to evaluating machine translation (MT) quality and also to addressing the so far unanswered question when and how to use MT in professional translation workflows. This paper describes the corpus developed as a result of a detailed large scale human evaluation consisting of three tightly connected tasks: ranking, error classification and post-editing.

Significant breakthroughs in machine translation only seem possible if human translators are taken into the loop. While automatic evaluation and scoring mechanisms such as BLEU have enabled the fast development of systems, it is not clear how systems can meet real-world (quality) requirements in industrial translation scenarios today. The taraXÜ project paves the way for wide usage of hybrid machine translation outputs through various feedback loops in system development. In a consortium of research and industry partners, the project integrates human translators into the development process for rating and post-editing of machine translation outputs thus collecting feedback for possible improvements.

Several studies indicate that the level of predicate-argument structure is relevant for modeling prevalent phenomena in current textual entailment corpora. Although large resources like FrameNet have recently become available, attempts to integrate this type of information into a system for textual entailment did not confirm the expected gain in performance. The reasons for this are not fully obvious; candidates include FrameNet’s restricted coverage, limitations of semantic parsers, or insufficient modeling of FrameNet information. To enable further insight on this issue, in this paper we present FATE (FrameNet-Annotated Textual Entailment), a manually crafted, fully reliable frame-annotated RTE corpus. The annotation has been carried out over the 800 pairs of the RTE-2 test set. This dataset offers a safe basis for RTE systems to experiment, and enables researchers to develop clearer ideas on how to effectively integrate frame knowledge in semantic inferenence tasks like recognizing textual entailment. We describe and present statistics over the adopted annotation, which introduces a new schema based on full-text annotation of so called relevant frame evoking elements.

This paper describes the SALSA corpus, a large German corpus manually annotated with manual role-semantic annotation, based on the syntactically annotated TIGER newspaper corpus. The first release, comprising about 20,000 annotated predicate instances (about half the TIGER corpus), is scheduled for mid-2006. In this paper we discuss the annotation framework (frame semantics) and its cross-lingual applicability, problems arising from exhaustive annotation, strategies for quality control, and possible applications.

In this paper, we describe the SALTO tool. It was originally developed for the annotation of semantic roles in the frame semantics paradigm, but can be used for graphical annotation of treebanks with general relational information in a simple drag-and-drop fashion. The tool additionally supports corpus management and quality control.