Named entity linking (NEL) in news is a challenging endeavour due to the frequency of unseen and emerging entities, which necessitates the use of unsupervised or zero-shot methods. However, such methods tend to come with caveats, such as no integration of suitable knowledge bases (like Wikidata) for emerging entities, a lack of scalability, and poor interpretability. Here, we consider person disambiguation in Quotebank, a massive corpus of speaker-attributed quotations from the news, and investigate the suitability of intuitive, lightweight, and scalable heuristics for NEL in web-scale corpora. Our best performing heuristic disambiguates 94% and 63% of the mentions on Quotebank and the AIDA-CoNLL benchmark, respectively. Additionally, the proposed heuristics compare favourably to the state-of-the-art unsupervised and zero-shot methods, Eigenthemes and mGENRE, respectively, thereby serving as strong baselines for unsupervised and zero-shot entity linking.

The awareness and mitigation of biases are of fundamental importance for the fair and transparent use of contextual language models, yet they crucially depend on the accurate detection of biases as a precursor. Consequently, numerous bias detection methods have been proposed, which vary in their approach, the considered type of bias, and the data used for evaluation. However, while most detection methods are derived from the word embedding association test for static word embeddings, the reported results are heterogeneous, inconsistent, and ultimately inconclusive. To address this issue, we conduct a rigorous analysis and comparison of bias detection methods for contextual language models. Our results show that minor design and implementation decisions (or errors) have a substantial and often significant impact on the derived bias scores. Overall, we find the state of the field to be both worse than previously acknowledged due to systematic and propagated errors in implementations, yet better than anticipated since divergent results in the literature homogenize after accounting for implementation errors. Based on our findings, we conclude with a discussion of paths towards more robust and consistent bias detection methods.

Geographic information extraction from textual data sources, called geoparsing, is a key task in text processing and central to subsequent spatial analysis approaches. Several geoparsers are available that support this task, each with its own (often limited or specialized) gazetteer and its own approaches to toponym detection and resolution. In this demonstration paper, we present HeidelPlace, an extensible framework in support of geoparsing. Key features of HeidelPlace include a generic gazetteer model that supports the integration of place information from different knowledge bases, and a pipeline approach that enables an effective combination of diverse modules tailored to specific geoparsing tasks. This makes HeidelPlace a valuable tool for testing and evaluating different gazetteer sources and geoparsing methods. In the demonstration, we show how to set up a geoparsing workflow with HeidelPlace and how it can be used to compare and consolidate the output of different geoparsing approaches.