Previous studies have highlighted the importance of vaccination as an effective strategy to control the transmission of the COVID-19 virus. It is crucial for policymakers to have a comprehensive understanding of the public's stance towards vaccination on a large scale. However, attitudes towards COVID-19 vaccination, such as pro-vaccine or vaccine hesitancy, have evolved over time on social media. Thus, it is necessary to account for possible temporal shifts when analysing these stances. This study aims to examine the impact of temporal concept drift on stance detection towards COVID-19 vaccination on Twitter. To this end, we evaluate a range of transformer-based models using chronological (split the training, validation and testing sets in the order of time) and random splits (randomly split these three sets) of social media data. Our findings demonstrate significant discrepancies in model performance when comparing random and chronological splits across all monolingual and multilingual datasets. Chronological splits significantly reduce the accuracy of stance classification. Therefore, real-world stance detection approaches need to be further refined to incorporate temporal factors as a key consideration.

Temporal concept drift refers to the problem of data changing over time. In NLP, that would entail that language (e.g. new expressions, meaning shifts) and factual knowledge (e.g. new concepts, updated facts) evolve over time. Focusing on the latter, we benchmark $11$ pretrained masked language models (MLMs) on a series of tests designed to evaluate the effect of temporal concept drift, as it is crucial that widely used language models remain up-to-date with the ever-evolving factual updates of the real world. Specifically, we provide a holistic framework that (1) dynamically creates temporal test sets of any time granularity (e.g. month, quarter, year) of factual data from Wikidata, (2) constructs fine-grained splits of tests (e.g. updated, new, unchanged facts) to ensure comprehensive analysis, and (3) evaluates MLMs in three distinct ways (single-token probing, multi-token generation, MLM scoring). In contrast to prior work, our framework aims to unveil how robust an MLM is over time and thus to provide a signal in case it has become outdated, by leveraging multiple views of evaluation.

New events emerge over time influencing the topics of rumors in social media. Current rumor detection benchmarks use random splits as training, development and test sets which typically results in topical overlaps. Consequently, models trained on random splits may not perform well on rumor classification on previously unseen topics due to the temporal concept drift. In this paper, we provide a re-evaluation of classification models on four popular rumor detection benchmarks considering chronological instead of random splits. Our experimental results show that the use of random splits can significantly overestimate predictive performance across all datasets and models. Therefore, we suggest that rumor detection models should always be evaluated using chronological splits for minimizing topical overlaps.

This research explores temporal concept drift and temporal alignment in knowledge organization systems (KOS). A comparative analysis is pursued using the 1910 Library of Congress Subject Headings, 2020 FAST Topical, and automatic indexing. The use case involves a sample of 90 nineteenth-century Encyclopedia Britannica entries. The entries were indexed using two approaches: 1) full-text indexing; 2) Named Entity Recognition was performed upon the entries with Stanza, Stanford's NLP toolkit, and entities were automatically indexed with the Helping Interdisciplinary Vocabulary application (HIVE), using both 1910 LCSH and FAST Topical. The analysis focused on three goals: 1) identifying results that were exclusive to the 1910 LCSH output; 2) identifying terms in the exclusive set that have been deprecated from the contemporary LCSH, demonstrating temporal concept drift; and 3) exploring the historical significance of these deprecated terms. Results confirm that historical vocabularies can be used to generate anachronistic subject headings representing conceptual drift across time in KOS and historical resources. A methodological contribution is made demonstrating how to study changes in KOS over time and improve the contextualization of historical humanities resources.