A Deep Metric Learning Approach to Account Linking

Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 5275–5287 June 6–11, 2021. ©2021 Association for Computational Linguistics 5275 A Deep Metric Learning Approach to Account Linking Aleem Khan, Elizabeth Fleming, Noah Schofield, Marcus Bishop, and Nicholas Andrews Human Language Technology Center of Excellence Johns Hopkins University {akhan141,eflemin9,nschofi1,marcus.bishop,noa}@jhu.edu Abstract We consider the task of linking social media accounts that belong to the same author in an automated fashion on the basis of the con- tent and metadata of their corresponding docu- ment streams. We focus on learning an embed- ding that maps variable-sized samples of user activity—ranging from single posts to entire months of activity—to a vector space, where samples by the same author map to nearby points. The approach does not require human- annotated data for training purposes, which al- lows us to leverage large amounts of social media content. The proposed model outper- forms several competitive baselines under a novel evaluation framework modeled after es- tablished recognition benchmarks in other do- mains. Our method achieves high linking accu- racy, even with small samples from accounts not seen at training time, a prerequisite for practical applications of the proposed linking framework. 1 Introduction The scale and anonymity of social media pose systematic challenges for manual moderation ef- forts (Pennycook et al., 2020; Broniatowski et al., 2018). These challenges have motivated the de- velopment of automated methods to identify abu- sive content, such as Davidson et al. (2017), which considers automatically classifying hate speech, or Alvari et al. (2019), which deals with detecting violent extremism, both in the Twitter domain. However, automatic moderation remains a diffi- cult problem. Indeed, existing methods based on hand-constructed resources such as keyword lists may fail to adapt to novel trends (Corbett-Davies and Goel, 2018) whereas automatic methods based on statistics of large corpora may exhibit harmful biases (Caliskan et al., 2017). Additionally, individ- ual posts may fail to contain sufficient information to reliably identify them as harmful. This work considers account-level moderation. Specifically, we consider the problem of determin- ing whether two document streams share the same author, based on samples from those streams rather than individual documents. This capability has numerous applications, such as detecting users at- tempting to circumvent account bans, identifying sockpuppet accounts, and detecting coordinated disinformation campaigns involving multiple au- thors controlling multiple accounts. As a motivating application, we consider the enforcement of account bans on anonymous plat- forms, such as Reddit. Given a new account, the problem is to automatically identify whether it matches any previously banned account, which amounts to making binary decisions about whether pairs of accounts share the same author. Variations of this problem have been studied before. For ex- ample, Schwartz et al. (2013) learn a classifier to determine whether the author of a Twitter comment belongs to a small, closed set of authors. In con- trast, we are interested in an open-world setting, requiring binary decisions about arbitrary pairs of accounts. This introduces …

Content truncated to 500 words