We present a named entity annotation for the Danish Universal Dependencies treebank using the CoNLL-2003 annotation scheme: DaNE. It is the largest publicly available, Danish named entity gold annotation. We evaluate the quality of our annotations intrinsically by double annotating the entire treebank and extrinsically by comparing our annotations to a recently released named entity annotation of the validation and test sections of the Danish Universal Dependencies treebank. We benchmark the new resource by training and evaluating competitive architectures for supervised named entity recognition (NER), including FLAIR, monolingual (Danish) BERT and multilingual BERT. We explore cross-lingual transfer in multilingual BERT from five related languages in zero-shot and direct transfer setups, and we show that even with our modestly-sized training set, we improve Danish NER over a recent cross-lingual approach, as well as over zero-shot transfer from five related languages. Using multilingual BERT, we achieve higher performance by fine-tuning on both DaNE and a larger Bokmål (Norwegian) training set compared to only using DaNE. However, the highest performance isachieved by using a Danish BERT fine-tuned on DaNE. Our dataset enables improvements and applicability for Danish NER beyond cross-lingual methods. We employ a thorough error analysis of the predictions of the best models for seen and unseen entities, as well as their robustness on un-capitalized text. The annotated dataset and all the trained models are made publicly available.

This paper presents the process of compiling a model-agnostic similarity goal standard for evaluating Danish word embeddings based on human judgments made by 42 native speakers of Danish. Word embeddings resemble semantic similarity solely by distribution (meaning that word vectors do not reflect relatedness as differing from similarity), and we argue that this generalization poses a problem in most intrinsic evaluation scenarios. In order to be able to evaluate on both dimensions, our human-generated dataset is therefore designed to reflect the distinction between relatedness and similarity. The goal standard is applied for evaluating the “goodness” of six existing word embedding models for Danish, and it is discussed how a relatively low correlation can be explained by the fact that semantic similarity is substantially more challenging to model than relatedness, and that there seems to be a need for future human judgments to measure similarity in full context and along more than a single spectrum.