Hybrid Uncertainty Quantification for Selective Text Classification in Ambiguous Tasks

Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics Volume 1: Long Papers, pages 11659–11681 July 9-14, 2023 ©2023 Association for Computational Linguistics Hybrid Uncertainty Quantification for Selective Text Classification in Ambiguous Tasks Artem Vazhentsev1,2, Gleb Kuzmin1,4, Akim Tsvigun5,8, Alexander Panchenko2,1, Maxim Panov6, Mikhail Burtsev7, and Artem Shelmanov3 1AIRI, 2Skoltech, 3MBZUAI, 4FRC CSC RAS, 5AI Center NUST MISiS, 6TII, 7London Institute for Mathematical Sciences, 8Semrush {vazhentsev, kuzmin, panchenko}@airi.net maxim.panov@tii.ae artem.shelmanov@mbzuai.ac.ae Abstract Many text classification tasks are inherently ambiguous, which results in automatic sys- tems having a high risk of making mistakes, in spite of using advanced machine learning models. For example, toxicity detection in user- generated content is a subjective task, and no- tions of toxicity can be annotated according to a variety of definitions that can be in conflict with one another. Instead of relying solely on automatic solutions, moderation of the most difficult and ambiguous cases can be delegated to human workers. Potential mistakes in auto- mated classification can be identified by using uncertainty estimation (UE) techniques. Al- though UE is a rapidly growing field within natural language processing, we find that state- of-the-art UE methods estimate only epistemic uncertainty and show poor performance, or under-perform trivial methods for ambiguous tasks such as toxicity detection. We argue that in order to create robust uncertainty estimation methods for ambiguous tasks it is necessary to account also for aleatoric uncertainty. In this paper, we propose a new uncertainty estimation method that combines epistemic and aleatoric UE methods. We show that by using our hybrid method, we can outperform state-of-the-art UE methods for toxicity detection and other am- biguous text classification tasks1. 1 Introduction Many natural language processing (NLP) tasks are subjective and contain inherent ambiguity. For ex- ample, the notion of toxicity is inherently subjec- tive (Waseem, 2016) and can be defined in a num- ber of ways that may conflict with one another and differ according to the demographic that the meth- ods are applied to (Thylstrup and Waseem, 2020). For many datasets, implicit or ambiguous toxicity can comprise more than 90% of the labeled toxic 1The code for reproducing experiments is available online at https://github.com/AIRI-Institute/hybrid_ uncertainty_estimation content (Hartvigsen et al., 2022). Such ambiguity introduces a high risk of classification mistakes for machine learning (ML) models. Classification mis- takes for toxicity detection can result in the removal of legitimate non-toxic content on one hand, and the lack of sanction for toxic content, on the other. A common method for addressing this concern for content moderation is to abstain from predictions on ambiguous instances and process them with the help of human workers (Roberts, 2019). A classification task where some model predic- tions can be “rejected” is called selective classifica- tion (Geifman and El-Yaniv, 2017). The common approach to solving it is applying uncertainty es- timation (UE) techniques. UE is a field of ML that seeks to model the degree to which model predictions can be trusted by correlating model mistakes and performance. Better UE methods im- …

Content truncated to 500 words