We present CACTI, a masked autoencoding approach for imputing tabular data that leverages the structure in missingness patterns and contextual information. Our approach employs a novel median truncated copy masking training strategy that encourages the model to learn from empirical patterns of missingness while incorporating semantic relationships between features — captured by column names and text descriptions — to better represent feature dependence. These dual sources of inductive bias enable CACTIto outperform state-of-the-art methods — an average $R^2$ gain of 7.8\% over the next best method (13.4%, 6.1%, and 5.3% under missing not at random, at random and completely at random, respectively) — across a diverse range of datasets and missingness conditions. Our results highlight the value of leveraging dataset-specific contextual information and missingness patterns to enhance imputation performance.

Using Knowledge Graphs to understand noisy naturalistic data has gained significant prominence in recent years. In this paper, we apply Knowledge Graphs to a new dataset of tweets of an ideologically far-right Twitter network by sourcing tweet histories of users who discussed QAnon in the summer of 2018 [1]. We further develop a new method that arms topic models with relational information from Knowledge Graphs and apply the new technique to study this dataset. Our analysis shows that users do not form a monolithic belief or social network, but rather comprise many smaller interlinking communities which discuss unique key political events (e.g., the January 6thCapitol riots).