Online abusive content detection is an inherently difficult task. It has received considerable attention from academia, particularly within the computational linguistics community, and performance appears to have improved as the field has matured. However, considerable challenges and unaddressed frontiers remain, spanning technical, social and ethical dimensions. These issues constrain the performance, efficiency and generalizability of abusive content detection systems. In this article we delineate and clarify the main challenges and frontiers in the field, critically evaluate their implications and discuss solutions. We also highlight ways in which social scientific insights can advance research.

Social media provide access to behavioural data at an unprecedented scale and granularity. However, using these data to understand phenomena in a broader population is difficult due to their nonrepresentativeness and the bias of statistical inference tools towards dominant languages and groups. While demographic attribute inference could be used to mitigate such bias, current techniques are almost entirely monolingual and fail to work in a global environment. We address these challenges by combining multilingual demographic inference with post-stratification to create a more representative population sample. To learn demographic attributes, we create a new multimodal deep neural architecture for joint classification of age, gender, and organization-status of social media users that operates in 32 languages. This method substantially outperforms current state of the art while also reducing algorithmic bias. To correct for sampling biases, we propose fully interpretable multilevel regression methods that estimate inclusion probabilities from inferred joint population counts and ground-truth population counts. In a large experiment over multilingual heterogeneous European regions, we show that our demographic inference and bias correction together allow for more accurate estimates of populations and make a significant step towards representative social sensing in downstream applications with multilingual social media.

All authors contributed equally to this paper. AbstractThe movements of ideas and content between locations and languages are unquestionably crucial concerns to researchers of the information age, and Twitter has emerged as a central, global platform on which hundreds of millions of people share knowledge and information. A variety of research has attempted to harvest locational and linguistic metadata from tweets in order to understand important questions related to the 300 million tweets that flow through the platform each day. However, much of this work is carried out with only limited understandings of how best to work with the spatial and linguistic contexts in which the information was produced. Furthermore, standard, well-accepted practices have yet to emerge. As such, this paper studies the reliability of key methods used to determine language and location of content in Twitter. It compares three automated language identification packages to Twitter's user interface language setting and to a human coding of languages in order to identify common sources of disagreement. The paper also demonstrates that in many cases user-entered profile locations differ from the physical locations users are actually tweeting from. As such, these open-ended, user-generated, profile locations cannot be used as useful proxies for the physical locations from which information is published to Twitter.

Word embeddings are increasingly used for the automatic detection of semantic change; yet, a robust evaluation and systematic comparison of the choices involved has been lacking. We propose a new evaluation framework for semantic change detection and find that (i) using the whole time series is preferable over only comparing between the first and last time points; (ii) independently trained and aligned embeddings perform better than continuously trained embeddings for long time periods; and (iii) that the reference point for comparison matters. We also present an analysis of the changes detected on a large Twitter dataset spanning 5.5 years.

During COVID-19 concerns have heightened about the spread of aggressive and hateful language online, especially hostility directed against East Asia and East Asian people. We report on a new dataset and the creation of a machine learning classifier that categorizes social media posts from Twitter into four classes: Hostility against East Asia, Criticism of East Asia, Meta-discussions of East Asian prejudice, and Non-related. The classifier achieves a macro-F1 score of 0.83. We then conduct an in-depth ground-up error analysis and show that the model struggles with edge cases and ambiguous content. We provide the 20,000 tweet training dataset (annotated by experienced analysts), which also contains several secondary categories and additional flags. We also provide the 40,000 original annotations (before adjudication), the full codebook, annotations for COVID-19 relevance and East Asian relevance and stance for 1,000 hashtags, and the final model.