Recent advances in natural language processing (NLP) can be largely attributed to the advent of pre-trained language models such as BERT and RoBERTa. While these models demonstrate remarkable performance on general datasets, they can struggle in specialized domains such as medicine, where unique domain-specific terminologies, domain-specific abbreviations, and varying document structures are common. This paper explores strategies for adapting these models to domain-specific requirements, primarily through continuous pre-training on domain-specific data. We pre-trained several German medical language models on 2.4B tokens derived from translated public English medical data and 3B tokens of German clinical data. The resulting models were evaluated on various German downstream tasks, including named entity recognition (NER), multi-label classification, and extractive question answering. Our results suggest that models augmented by clinical and translation-based pre-training typically outperform general domain models in medical contexts. We conclude that continuous pre-training has demonstrated the ability to match or even exceed the performance of clinical models trained from scratch. Furthermore, pre-training on clinical data or leveraging translated texts have proven to be reliable methods for domain adaptation in medical NLP tasks.
Deep learning models have achieved high performance in medical applications, however, their adoption in clinical practice is hindered due to their black-box nature. Self-explainable models, like prototype-based models, can be especially beneficial as they are interpretable by design. However, if the learnt prototypes are of low quality then the prototype-based models are as good as black-box. Having high quality prototypes is a pre-requisite for a truly interpretable model. In this work, we propose a prototype evaluation framework for coherence (PEF-C) for quantitatively evaluating the quality of the prototypes based on domain knowledge. We show the use of PEF-C in the context of breast cancer prediction using mammography. Existing works on prototype-based models on breast cancer prediction using mammography have focused on improving the classification performance of prototype-based models compared to black-box models and have evaluated prototype quality through anecdotal evidence. We are the first to go beyond anecdotal evidence and evaluate the quality of the mammography prototypes systematically using our PEF-C. Specifically, we apply three state-of-the-art prototype-based models, ProtoPNet, BRAIxProtoPNet++ and PIP-Net on mammography images for breast cancer prediction and evaluate these models w.r.t. i) classification performance, and ii) quality of the prototypes, on three public datasets. Our results show that prototype-based models are competitive with black-box models in terms of classification performance, and achieve a higher score in detecting ROIs. However, the quality of the prototypes are not yet sufficient and can be improved in aspects of relevance, purity and learning a variety of prototypes. We call the XAI community to systematically evaluate the quality of the prototypes to check their true usability in high stake decisions and improve such models further.
Information from neuroimaging examinations (CT, MRI) is increasingly used to support diagnoses of dementia, e.g., Alzheimer's disease. While current clinical practice is mainly based on visual inspection and feature engineering, Deep Learning approaches can be used to automate the analysis and to discover new image-biomarkers. Part-prototype neural networks (PP-NN) are an alternative to standard blackbox models, and have shown promising results in general computer vision. PP-NN's base their reasoning on prototypical image regions that are learned fully unsupervised, and combined with a simple-to-understand decision layer. We present PIPNet3D, a PP-NN for volumetric images. We apply PIPNet3D to the clinical case study of Alzheimer's Disease diagnosis from structural Magnetic Resonance Imaging (sMRI). We assess the quality of prototypes under a systematic evaluation framework, propose new metrics to evaluate brain prototypes and perform an evaluation with domain experts. Our results show that PIPNet3D is an interpretable, compact model for Alzheimer's diagnosis with its reasoning well aligned to medical domain knowledge. Notably, PIPNet3D achieves the same accuracy as its blackbox counterpart; and removing the remaining clinically irrelevant prototypes from its decision process does not decrease predictive performance.
We present a detailed replication study of the BASS framework, an abstractive summarization system based on the notion of Unified Semantic Graphs. Our investigation includes challenges in replicating key components and an ablation study to systematically isolate error sources rooted in replicating novel components. Our findings reveal discrepancies in performance compared to the original work. We highlight the significance of paying careful attention even to reasonably omitted details for replicating advanced frameworks like BASS, and emphasize key practices for writing replicable papers.
Factual knowledge encoded in Pre-trained Language Models (PLMs) enriches their representations and justifies their use as knowledge bases. Previous work has focused on probing PLMs for factual knowledge by measuring how often they can correctly predict an object entity given a subject and a relation, and improving fact retrieval by optimizing the prompts used for querying PLMs. In this work, we consider a complementary aspect, namely the coherency of factual knowledge in PLMs, i.e., how often can PLMs predict the subject entity given its initial prediction of the object entity. This goes beyond evaluating how much PLMs know, and focuses on the internal state of knowledge inside them. Our results indicate that PLMs have low coherency using manually written, optimized and paraphrased prompts, but including an evidence paragraph leads to substantial improvement. This shows that PLMs fail to model inverse relations and need further enhancements to be able to handle retrieving facts from their parameters in a coherent manner, and to be considered as knowledge bases.
Text simplification aims to make technical texts more accessible to laypeople but often results in deletion of information and vagueness. This work proposes InfoLossQA, a framework to characterize and recover simplification-induced information loss in form of question-and-answer (QA) pairs. Building on the theory of Question Under Discussion, the QA pairs are designed to help readers deepen their knowledge of a text. We conduct a range of experiments with this framework. First, we collect a dataset of 1,000 linguist-curated QA pairs derived from 104 LLM simplifications of scientific abstracts of medical studies. Our analyses of this data reveal that information loss occurs frequently, and that the QA pairs give a high-level overview of what information was lost. Second, we devise two methods for this task: end-to-end prompting of open-source and commercial language models, and a natural language inference pipeline. With a novel evaluation framework considering the correctness of QA pairs and their linguistic suitability, our expert evaluation reveals that models struggle to reliably identify information loss and applying similar standards as humans at what constitutes information loss.
In industry, Bayesian optimization (BO) is widely applied in the human-AI collaborative parameter tuning of cyber-physical systems. However, BO's solutions may deviate from human experts' actual goal due to approximation errors and simplified objectives, requiring subsequent tuning. The black-box nature of BO limits the collaborative tuning process because the expert does not trust the BO recommendations. Current explainable AI (XAI) methods are not tailored for optimization and thus fall short of addressing this gap. To bridge this gap, we propose TNTRules (TUNE-NOTUNE Rules), a post-hoc, rule-based explainability method that produces high quality explanations through multiobjective optimization. Our evaluation of benchmark optimization problems and real-world hyperparameter optimization tasks demonstrates TNTRules' superiority over state-of-the-art XAI methods in generating high quality explanations. This work contributes to the intersection of BO and XAI, providing interpretable optimization techniques for real-world applications.
Third-generation artificial neural networks, Spiking Neural Networks (SNNs), can be efficiently implemented on hardware. Their implementation on neuromorphic chips opens a broad range of applications, such as machine learning-based autonomous control and intelligent biomedical devices. In critical applications, however, insight into the reasoning of SNNs is important, thus SNNs need to be equipped with the ability to explain how decisions are reached. We present \textit{Temporal Spike Attribution} (TSA), a local explanation method for SNNs. To compute the explanation, we aggregate all information available in model-internal variables: spike times and model weights. We evaluate TSA on artificial and real-world time series data and measure explanation quality w.r.t. multiple quantitative criteria. We find that TSA correctly identifies a small subset of input features relevant to the decision (i.e., is output-complete and compact) and generates similar explanations for similar inputs (i.e., is continuous). Further, our experiments show that incorporating the notion of \emph{absent} spikes improves explanation quality. Our work can serve as a starting point for explainable SNNs, with future implementations on hardware yielding not only predictions but also explanations in a broad range of application scenarios. Source code is available at https://github.com/ElisaNguyen/tsa-explanations.
Pre-trained Language Models (PLMs) are trained on vast unlabeled data, rich in world knowledge. This fact has sparked the interest of the community in quantifying the amount of factual knowledge present in PLMs, as this explains their performance on downstream tasks, and potentially justifies their use as knowledge bases. In this work, we survey methods and datasets that are used to probe PLMs for factual knowledge. Our contributions are: (1) We propose a categorization scheme for factual probing methods that is based on how their inputs, outputs and the probed PLMs are adapted; (2) We provide an overview of the datasets used for factual probing; (3) We synthesize insights about knowledge retention and prompt optimization in PLMs, analyze obstacles to adopting PLMs as knowledge bases and outline directions for future work.
Automatic methods for early detection of breast cancer on mammography can significantly decrease mortality. Broad uptake of those methods in hospitals is currently hindered because the methods have too many constraints. They assume annotations available for single images or even regions-of-interest (ROIs), and a fixed number of images per patient. Both assumptions do not hold in a general hospital setting. Relaxing those assumptions results in a weakly supervised learning setting, where labels are available per case, but not for individual images or ROIs. Not all images taken for a patient contain malignant regions and the malignant ROIs cover only a tiny part of an image, whereas most image regions represent benign tissue. In this work, we investigate a two-level multi-instance learning (MIL) approach for case-level breast cancer prediction on two public datasets (1.6k and 5k cases) and an in-house dataset of 21k cases. Observing that breast cancer is usually only present in one side, while images of both breasts are taken as a precaution, we propose a domain-specific MIL pooling variant. We show that two-level MIL can be applied in realistic clinical settings where only case labels, and a variable number of images per patient are available. Data in realistic settings scales with continuous patient intake, while manual annotation efforts do not. Hence, research should focus in particular on unsupervised ROI extraction, in order to improve breast cancer prediction for all patients.