Research probing the language comprehension of visio-linguistic models has gained traction due to their remarkable performance on various tasks. We introduce EViL-Probe, a composite benchmark that processes existing probing datasets into a unified format and reorganizes them based on the linguistic categories they probe. On top of the commonly used negative probes, this benchmark introduces positive probes to more rigorously test the robustness of models. Since the language side alone may introduce a bias models could exploit in solving the probes, we estimate the difficulty of the individual subsets with a language-only baseline. Using the benchmark to probe a set of state-of-the-art visio-linguistic models sheds light on how sensitive they are to the different linguistic categories. Results show that the benchmark is challenging for all models we probe, as their performance is around the chance baseline for many of the categories. The only category all models are able to handle relatively well are nouns. Additionally, models that use a Vision Transformer to process the images are also somewhat robust against probes targeting color and image type. Among these models, our enrichment of EViL-Probe with positive probes helps further discriminate performance, showing BLIP to be the overall best-performing model.

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.

Developmental stages are a linguistic concept claiming that language learning, despite its large inter-individual variance, generally progresses in an ordered, step-like manner. At the core of research has been the acquisition of verb placement by learners, as conceptualized within Processability Theory (Pienemann, 1989). The computational implementation of a system detecting developmental stages is a prerequisite for an automated analysis of L2 language development. However, such an implementation faces two main challenges. The first is the lack of a fully fleshed out, coherent linguistic specification of the stages. The second concerns the translation of the linguistic specification into computational procedures that can extract clauses from learner-produced text and assign them to a developmental stage based on verb placement. Our contribution provides the necessary linguistic specification of the stages as well as detaiiled discussion and recommendations regarding computational implementation.

With the recent emergence of powerful visio-linguistic models comes the question of how fine-grained their multi-modal understanding is. This has lead to the release of several probing datasets. Results point towards models having trouble with prepositions and verbs, but being relatively robust when it comes to color.To gauge how deep this understanding goes, we compile a comprehensive probing dataset to systematically test multi-modal alignment around color. We demonstrate how human perception influences descriptions of color and pay special attention to the extent to which this is reflected within the predictions of a visio-linguistic model. Probing a set of models with diverse properties with our benchmark confirms the superiority of models that do not rely on pre-extracted image features, and demonstrates that augmentation with too much noisy pre-training data can produce an inferior model. While the benchmark remains challenging for all models we test, the overall result pattern suggests well-founded alignment of color terms with hues. Analyses do however reveal uncertainty regarding the boundaries between neighboring color terms.

Exploiting social media to spread hate has tremendously increased over the years. Lately, multi-modal hateful content such as memes has drawn relatively more traction than uni-modal content. Moreover, the availability of implicit content payloads makes them fairly challenging to be detected by existing hateful meme detection systems. In this paper, we present a use case study to analyze such systems’ vulnerabilities against external adversarial attacks. We find that even very simple perturbations in uni-modal and multi-modal settings performed by humans with little knowledge about the model can make the existing detection models highly vulnerable. Empirically, we find a noticeable performance drop of as high as 10% in the macro-F1 score for certain attacks. As a remedy, we attempt to boost the model’s robustness using contrastive learning as well as an adversarial training-based method - VILLA. Using an ensemble of the above two approaches, in two of our high resolution datasets, we are able to (re)gain back the performance to a large extent for certain attacks. We believe that ours is a first step toward addressing this crucial problem in an adversarial setting and would inspire more such investigations in the future.

In this work, we describe the findings of the 'WisPerMed' team from their participation in Track 1 (Contextualized Medication Event Extraction) of the n2c2 2022 challenge. We tackle two tasks: (i) medication extraction, which involves extracting all mentions of medications from the clinical notes, and (ii) event classification, which involves classifying the medication mentions based on whether a change in the medication has been discussed. To address the long lengths of clinical texts, which often exceed the maximum token length that models based on the transformer-architecture can handle, various approaches, such as the use of ClinicalBERT with a sliding window approach and Longformer-based models, are employed. In addition, domain adaptation through masked language modeling and preprocessing steps such as sentence splitting are utilized to improve model performance. Since both tasks were treated as named entity recognition (NER) problems, a sanity check was performed in the second release to eliminate possible weaknesses in the medication detection itself. This check used the medication spans to remove false positive predictions and replace missed tokens with the highest softmax probability of the disposition types. The effectiveness of these approaches is evaluated through multiple submissions to the tasks, as well as with post-challenge results, with a focus on the DeBERTa v3 model and its disentangled attention mechanism. Results show that the DeBERTa v3 model performs well in both the NER task and the event classification task.

Spellchecking text written by language learners is especially challenging because errors made by learners differ both quantitatively and qualitatively from errors made by already proficient learners. We introduce LeSpell, a multi-lingual (English, German, Italian, and Czech) evaluation data set of spelling mistakes in context that we compiled from seven underlying learner corpora. Our experiments show that existing spellcheckers do not work well with learner data. Thus, we introduce a highly customizable spellchecking component for the DKPro architecture, which improves performance in many settings.

Automatic content scoring systems are widely used on short answer tasks to save human effort.However, the use of these systems can invite cheating strategies, such as students writing irrelevant answers in the hopes of gaining at least partial credit.We generate adversarial answers for benchmark content scoring datasets based on different methods of increasing sophistication and show that even simple methods lead to a surprising decrease in content scoring performance.As an extreme example, up to 60% of adversarial answers generated from random shuffling of words in real answers are accepted by a state-of-the-art scoring system.In addition to analyzing the vulnerabilities of content scoring systems, we examine countermeasures such as adversarial training and show that these measures improve system robustness against adversarial answers considerably but do not suffice to completely solve the problem.

Handwriting is still the preferred writing mode in the classroom, but it makes introducing intelligent tutoring systems much harder. A major building block within these systems are scoring algorithms that assess whether the content of free-text student answers is correct or not. We collect a dataset of handwritten student answers which we make publicly available to foster future research in the field. We use a state-of-the-art system for handwriting recognition containing a line and word segmentation module and a neural recognition model. We find that scoring performance on handwritten answers is significantly worse compared to typed answers due to recognition errors. Using a simple postprocessing step based on a unigram language model is already sufficient to achieve a performance very close to that on typed input.

In this paper, we present a methodology for decomposing and comparing multiple meaning relations (paraphrasing, textual entailment, contradiction, and specificity). The methodology includes SHARel - a new typology that consists of 26 linguistic and 8 reason-based categories. We use the typology to annotate a corpus of 520 sentence pairs in English and we demonstrate that unlike previous typologies, SHARel can be applied to all relations of interest with a high inter-annotator agreement. We analyze and compare the frequency and distribution of the linguistic and reason-based phenomena involved in paraphrasing, textual entailment, contradiction, and specificity. This comparison allows for a much more in-depth analysis of the workings of the individual relations and the way they interact and compare with each other. We release all resources (typology, annotation guidelines, and annotated corpus) to the community.

A recent study by Plank et al. (2016) found that LSTM-based PoS taggers considerably improve over the current state-of-theart when evaluated on the corpora of the Universal Dependencies project that use a coarse-grained tagset.We replicate this study using a fresh collection of 27 corpora of 21 languages that are annotated with fine-grained tagsets of varying size.Our replication confirms the result in general, and we additionally find that the advantage of LSTMs is even bigger for larger tagsets.However, we also find that for the very large tagsets of morphologically rich languages, hand-crafted morphological lexicons are still necessary to reach state-of-the-art performance.

We propose a new approach to PoS tagging where in a first step, we assign a coarse-grained tag corresponding to the main syntactic category. Based on this high-precision decision, in the second step we utilize specially trained fine-grained models with heavily reduced decision complexity. By analyzing the system under oracle conditions, we show that there is a quite large potential for significantly outperforming a competitive baseline. When we take error-propagation from the coarse-grained tagging into account, our approach is on par with the state of the art. Our approach also allows tailoring the tagger towards recognizing single word classes which are of interest e.g. for researchers searching for specific phenomena in large corpora. In a case study, we significantly outperform a standard model that also makes use of the same optimizations.

The lack of a sufficient amount of data tailored for a task is a well-recognized problem for many statistical NLP methods. In this paper, we explore whether data sparsity can be successfully tackled when classifying language proficiency levels in the domain of learner-written output texts. We aim at overcoming data sparsity by incorporating knowledge in the trained model from another domain consisting of input texts written by teaching professionals for learners. We compare different domain adaptation techniques and find that a weighted combination of the two types of data performs best, which can even rival systems based on considerably larger amounts of in-domain data. Moreover, we show that normalizing errors in learners’ texts can substantially improve classification when level-annotated in-domain data is not available.

Language proficiency tests are used to evaluate and compare the progress of language learners. We present an approach for automatic difficulty prediction of C-tests that performs on par with human experts. On the basis of detailed analysis of newly collected data, we develop a model for C-test difficulty introducing four dimensions: solution difficulty, candidate ambiguity, inter-gap dependency, and paragraph difficulty. We show that cues from all four dimensions contribute to C-test difficulty.

Abstract In this article, we present a comprehensive study aimed at computing semantic relatedness of word pairs. We analyze the performance of a large number of semantic relatedness measures proposed in the literature with respect to different experimental conditions, such as (i) the datasets employed, (ii) the language (English or German), (iii) the underlying knowledge source, and (iv) the evaluation task (computing scores of semantic relatedness, ranking word pairs, solving word choice problems). To our knowledge, this study is the first to systematically analyze semantic relatedness on a large number of datasets with different properties, while emphasizing the role of the knowledge source compiled either by the ‘wisdom of linguists’ (i.e., classical wordnets) or by the ‘wisdom of crowds’ (i.e., collaboratively constructed knowledge sources like Wikipedia). The article discusses benefits and drawbacks of different approaches to evaluating semantic relatedness. We show that results should be interpreted carefully to evaluate particular aspects of semantic relatedness. For the first time, we employ a vector based measure of semantic relatedness, relying on a concept space built from documents, to the first paragraph of Wikipedia articles, to English WordNet glosses, and to GermaNet based pseudo glosses. Contrary to previous research (Strube and Ponzetto 2006; Gabrilovich and Markovitch 2007; Zeschet al. 2007), we find that ‘wisdom of crowds’ based resources are not superior to ‘wisdom of linguists’ based resources. We also find that using the first paragraph of a Wikipedia article as opposed to the whole article leads to better precision, but decreases recall. Finally, we present two systems that were developed to aid the experiments presented herein and are freely available1for research purposes: (i) DEXTRACT, a software to semi-automatically construct corpus-driven semantic relatedness datasets, and (ii) JWPL, a Java-based high-performance Wikipedia Application Programming Interface (API) for building natural language processing (NLP) applications.