Jacob Aarup Dalsgaard

Anders Giovanni Møller, Jacob Aarup Dalsgaard, Arianna Pera et al.

In the realm of Computational Social Science (CSS), practitioners often navigate complex, low-resource domains and face the costly and time-intensive challenges of acquiring and annotating data. We aim to establish a set of guidelines to address such challenges, comparing the use of human-labeled data with synthetically generated data from GPT-4 and Llama-2 in ten distinct CSS classification tasks of varying complexity. Additionally, we examine the impact of training data sizes on performance. Our findings reveal that models trained on human-labeled data consistently exhibit superior or comparable performance compared to their synthetically augmented counterparts. Nevertheless, synthetic augmentation proves beneficial, particularly in improving performance on rare classes within multi-class tasks. Furthermore, we leverage GPT-4 and Llama-2 for zero-shot classification and find that, while they generally display strong performance, they often fall short when compared to specialized classifiers trained on moderately sized training sets.

Jacob Aarup Dalsgaard, Filipi Nascimento Silva, Jin AI

Funding acknowledgments in scholarly publications provide large-scale trace data on organizations that support scientific research. We present a dataset for linking global science funding organizations to research publications by systematically disambiguating unique funding acknowledgment strings extracted from publication metadata. Funder names are matched to standardized organizational identifiers using a multi-stage pipeline that combines lexical normalization, similarity-based clustering, rule-based matching, named entity recognition assistance, and manual validation. The resulting dataset links 1.9 million unique funder strings to canonical organization identifiers and records match types and unresolved cases to support transparency. Technical validation includes paper-level comparisons across bibliometric sources and manual verification against full-text acknowledgment sections, with reported recall and precision metrics. This dataset supports analyses of funding flows, institutional funding portfolios, regional representation, and concentration patterns in the global research system.

Leon Strømberg-Derczynski, Manuel R. Ciosici, Rebekah Baglini et al.

Danish language technology has been hindered by a lack of broad-coverage corpora at the scale modern NLP prefers. This paper describes the Danish Gigaword Corpus, the result of a focused effort to provide a diverse and freely-available one billion word corpus of Danish text. The Danish Gigaword corpus covers a wide array of time periods, domains, speakers' socio-economic status, and Danish dialects.