Detecting Vaccine Skepticism on Twitter Using Heterogeneous Information Networks

Identifying social media users who are skeptical of the COVID-19 vaccine is an important step in understanding and refuting negative stance taking on vaccines. While previous work on Twitter data places individual messages or whole communities as their focus, this paper aims to detect stance at the user level. We develop a system that classifies Dutch Twitter users, incorporating not only the texts that users produce, but also their actions in the form of following and retweeting. These heterogeneous data are modelled in a graph structure. Graph Convolutional Networks are trained to learn whether user nodes belong to the skeptical or non-skeptical group. Results show that all types of information are used by the model, and that especially user biographies, follows and retweets improve the predictions. On a test set of unseen users, performance declines somewhat, which is expected considering these users tweeted less and had fewer connections in the graph on average. To consider multiple degrees of vaccine skepticism, the test set was annotated with more fine-grained labels and the model was repurposed to do multiclass classification. While the model trained on binary labels was unsuited for this additional task, heterogeneous information networks were found useful to both accurately model and visualize complex user behaviors. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.

BART for knowledge grounded conversations

Transformers have shown incredible capabilities for conversation modeling, however, they store factual knowledge in their learned parameters, which is costly to update with new knowledge since it requires retraining. Models trained before the Coronavirus pandemic do not know about COVID-19. In this paper, we investigate how a BART model can be adapted to a knowledge grounded conversational setup. We introduce the notion of key and query tokens to retrieve knowledge stored in an external database, that can easily be updated with new knowledge. As factual knowledge can hardly be reduced to a single sentence or vector, we allow the model to retrieve multiple sentences from the memory. Our analysis shows perplexity decreases with the number of passages retrieved from memory. Second, our analysis shows a shared encoder for knowledge retrieval, and conversation understanding reduces the model size and perform as well as a specialized module. © 2020 Copyright held by the owner/author(s).