The extraction of valuable information from the vast amount of digital data available today has become increasingly important, making Named Entity Recognition models an essential component of information extraction tasks. This emphasizes the importance of understanding the factors that can compromise the performance of these models. Many studies have examined the impact of data annotation errors on NER models, leaving the broader implication of overall data quality on these models unexplored. In this work, we evaluate the robustness of three prominent NER models on datasets with varying amounts of textual noise types. The results show that as the noise in the dataset increases, model performance declines, with a minor impact for some noise types and a significant drop in performance for others. The findings of this research can be used as a foundation for building robust NER systems by enhancing dataset quality beforehand.

Differential diagnosis (DDx) is vital for physicians and challenging due to the existence of numerous diseases and their complex symptoms. Model training for this task is generally hindered by limited data access due to privacy concerns. To address this, we present DDxGym, a specialized OpenAI Gym environment for clinical differential diagnosis. DDxGym formulates DDx as a natural-language-based reinforcement learning (RL) problem, where agents emulate medical professionals, selecting examinations and treatments for patients with randomly sampled diseases. This RL environment utilizes data labeled from online resources, evaluated by medical professionals for accuracy. Transformers, while effective for encoding text in DDxGym, are unstable in online RL. For that reason we propose a novel training method using an auxiliary masked language modeling objective for policy optimization, resulting in model stabilization and significant performance improvement over strong baselines. Following this approach, our agent effectively navigates large action spaces and identifies universally applicable actions. All data, environment details, and implementation, including experiment reproduction code, are made publicly available.

With the rise of research on toxic comment classification, more and more annotated datasets have been released. The wide variety of the task (different languages, different labeling processes and schemes) has led to a large amount of heterogeneous datasets that can be used for training and testing very specific settings. Despite recent efforts to create web pages that provide an overview, most publications still use only a single dataset. They are not stored in one central database, they come in many different data formats and it is difficult to interpret their class labels and how to reuse these labels in other projects. To overcome these issues, we present a collection of more than thirty datasets in the form of a software tool that automatizes downloading and processing of the data and presents them in a unified data format that also offers a mapping of compatible class labels. Another advantage of that tool is that it gives an overview of properties of available datasets, such as different languages, platforms, and class labels to make it easier to select suitable training and test data.

Current document embeddings require large training corpora but fail to learn high-quality representations when confronted with a small number of domain-specific documents and rare terms. Further, they transform each document into a single embedding vector, making it hard to capture different notions of document similarity or explain why two documents are considered similar. In this work, we propose our Faceted Domain Encoder, a novel approach to learn multifaceted embeddings for domain-specific documents. It is based on a Siamese neural network architecture and leverages knowledge graphs to further enhance the embeddings even if only a few training samples are available. The model identifies different types of domain knowledge and encodes them into separate dimensions of the embedding, thereby enabling multiple ways of finding and comparing related documents in the vector space. We evaluate our approach on two benchmark datasets and find that it achieves the same embedding quality as state-of-the-art models while requiring only a tiny fraction of their training data. An interactive demo, our source code, and the evaluation datasets are available online: https://hpi.de/naumann/s/multifaceted-embeddings and a screencast is available on YouTube: https://youtu.be/HHcsX2clEwg

Named entity recognition (NER) is an important task that constitutes the basis for multiple downstream natural language processing tasks. Traditional machine learning approaches for NER rely on annotated corpora. However, these are only largely available for standard domains, e.g., news articles. Domain-specific NER often lacks annotated training data and therefore two options are of interest: expensive manual annotations or transfer learning. In this paper, we study a selection of cross-domain NER models and evaluate them for use in the art domain, particularly for recognizing artwork titles in digitized art-historic documents. For the evaluation of the models, we employ a variety of source domain datasets and analyze how each source domain dataset impacts the performance of the different models for our target domain. Additionally, we analyze the impact of the source domain’s entity types, looking for a better understanding of how the transfer learning models adapt different source entity types into our target entity types.

Languages evolve over time and the meaning of words can shift. Furthermore, individual words can have multiple senses. However, existing language models often only reflect one word sense per word and do not reflect semantic changes over time. While there are language models that can either model semantic change of words or multiple word senses, none of them cover both aspects simultaneously. We propose a novel force-directed graph layout algorithm to draw a network of frequently co-occurring words. In this way, we are able to use the drawn graph to visualize the evolution of word senses. In addition, we hope that jointly modeling semantic change and multiple senses of words results in improvements for the individual tasks.

Modern transformer-based models with hundreds of millions of parameters, such as BERT, achieve impressive results at text classification tasks. This also holds for aggression identification and offensive language detection, where deep learning approaches consistently outperform less complex models, such as decision trees. While the complex models fit training data well (low bias), they also come with an unwanted high variance. Especially when fine-tuning them on small datasets, the classification performance varies significantly for slightly different training data. To overcome the high variance and provide more robust predictions, we propose an ensemble of multiple fine-tuned BERT models based on bootstrap aggregating (bagging). In this paper, we describe such an ensemble system and present our submission to the shared tasks on aggression identification 2020 (team name: Julian). Our submission is the best-performing system for five out of six subtasks. For example, we achieve a weighted F1-score of 80.3% for task A on the test dataset of English social media posts. In our experiments, we compare different model configurations and vary the number of models used in the ensemble. We find that the F1-score drastically increases when ensembling up to 15 models, but the returns diminish for more models.

Many online discussion platforms use a content moderation process, where human moderators check user comments for offensive language and other rule violations. It is the moderator’s decision which comments to remove from the platform because of violations and which ones to keep. Research so far focused on automating this decision process in the form of supervised machine learning for a classification task. However, even with machine-learned models achieving better classification accuracy than human experts, there is still a reason why human moderators are preferred. In contrast to black-box models, such as neural networks, humans can give explanations for their decision to remove a comment. For example, they can point out which phrase in the comment is offensive or what subtype of offensiveness applies. In this paper, we analyze and compare four explanation methods for different offensive language classifiers: an interpretable machine learning model (naive Bayes), a model-agnostic explanation method (LIME), a model-based explanation method (LRP), and a self-explanatory model (LSTM with an attention mechanism). We evaluate these approaches with regard to their explanatory power and their ability to point out which words are most relevant for a classifier’s decision. We find that the more complex models achieve better classification accuracy while also providing better explanations than the simpler models.

Cultural heritage data plays a pivotal role in the understanding of human history and culture. A wealth of information is buried in art-historic archives which can be extracted via digitization and analysis. This information can facilitate search and browsing, help art historians to track the provenance of artworks and enable wider semantic text exploration for digital cultural resources. However, this information is contained in images of artworks, as well as textual descriptions or annotations accompanied with the images. During the digitization of such resources, the valuable associations between the images and texts are frequently lost. In this project description, we propose an approach to retrieve the associations between images and texts for artworks from art-historic archives. To this end, we use machine learning to generate text descriptions for the extracted images on the one hand, and to detect descriptive phrases and titles of images from the text on the other hand. Finally, we use embeddings to align both, the descriptions and the images.

The overwhelming success of the Web and mobile technologies has enabled millions to share their opinions publicly at any time. But the same success also endangers this freedom of speech due to closing down of participatory sites misused by individuals or interest groups. We propose to support manual moderation by proactively drawing the attention of our moderators to article discussions that most likely need their intervention. To this end, we predict which articles will receive a high number of comments. In contrast to existing work, we enrich the article with metadata, extract semantic and linguistic features, and exploit annotated data from a foreign language corpus. Our logistic regression model improves F1-scores by over 80% in comparison to state-of-the-art approaches.

Social media platforms allow users to share and discuss their opinions online. However, a minority of user posts is aggressive, thereby hinders respectful discussion, and — at an extreme level — is liable to prosecution. The automatic identification of such harmful posts is important, because it can support the costly manual moderation of online discussions. Further, the automation allows unprecedented analyses of discussion datasets that contain millions of posts. This system description paper presents our submission to the First Shared Task on Aggression Identification. We propose to augment the provided dataset to increase the number of labeled comments from 15,000 to 60,000. Thereby, we introduce linguistic variety into the dataset. As a consequence of the larger amount of training data, we are able to train a special deep neural net, which generalizes especially well to unseen data. To further boost the performance, we combine this neural net with three logistic regression classifiers trained on character and word n-grams, and hand-picked syntactic features. This ensemble is more robust than the individual single models. Our team named “Julian” achieves an F1-score of 60% on both English datasets, 63% on the Hindi Facebook dataset, and 38% on the Hindi Twitter dataset.

Comment sections of online news providers have enabled millions to share and discuss their opinions on news topics. Today, moderators ensure respectful and informative discussions by deleting not only insults, defamation, and hate speech, but also unverifiable facts. This process has to be transparent and comprehensive in order to keep the community engaged. Further, news providers have to make sure to not give the impression of censorship or dissemination of fake news. Yet manual moderation is very expensive and becomes more and more unfeasible with the increasing amount of comments. Hence, we propose a semi-automatic, holistic approach, which includes comment features but also their context, such as information about users and articles. For evaluation, we present experiments on a novel corpus of 3 million news comments annotated by a team of professional moderators.

Toxic comment classification has become an active research field with many recently proposed approaches. However, while these approaches address some of the task’s challenges others still remain unsolved and directions for further research are needed. To this end, we compare different deep learning and shallow approaches on a new, large comment dataset and propose an ensemble that outperforms all individual models. Further, we validate our findings on a second dataset. The results of the ensemble enable us to perform an extensive error analysis, which reveals open challenges for state-of-the-art methods and directions towards pending future research. These challenges include missing paradigmatic context and inconsistent dataset labels.

Reported speech in the form of direct and indirect reported speech is an important indicator of evidentiality in traditional newspaper texts, but also increasingly in the new media that rely heavily on citation and quotation of previous postings, as for instance in blogs or newsgroups. This paper details the basic processing steps for reported speech analysis and reports on performance of an implementation in form of a GATE resource.