Topics to Avoid: Demoting Latent Confounds in Text Classification

By Sachin Kumar, Shuly Wintner, Noah A. Smith, Yulia Tsvetkov. 2019

35 citations recorded by the source corpus.

Despite impressive performance on many text classification tasks, deep neural networks tend to learn frequent superficial patterns that are specific to the training data and do not always generalize well. In this work, we observe this limitation with respect to the task of native language identification. We find that standard text classifiers which perform well on the test set end up learning topical features which are confounds of the prediction task (e.g., if the input text mentions Sweden, the classifier predicts that the author’s native language is Swedish). We propose a method that represents the latent topical confounds and a model which “unlearns” confounding features by predicting both the label of the input text and the confound; but we train the two predictors adversarially in an alternating fashion to learn a text representation that predicts the correct label but is less prone to using information about the confound. We show that this model generalizes better and learns features that are indicative of the writing style rather than the content.

Subfields: Artificial Intelligence, Natural Language Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Natural Language Processing

Citations in the corpus

Language Generation Models Can Cause Harm: So What Can We Do About It? An Actionable Survey (2022, background): “box models, classifiers overfit to spurious artifacts (Gururangan et al., 2018; McCoy et al., 2019; Kumar et al., 2019) and amplify biases in their training data (Zhao et al.”
SELFEXPLAIN: A Self-Explaining Architecture for Neural Text Classifiers (2021, background): “There has been ample evidence that such models overﬁt to spurious artifacts (Gu-rurangan et al., 2018; McCoy et al., 2019; Kumar et al., 2019) and amplify biases in data (Zhao et al., 2017; Sun et al., 2019).”
Demoting Racial Bias in Hate Speech Detection (2020, methodology): “Unlike Kumar et al. (2019), we introduce a hyper-parameter α, which controls the balance between the two loss terms in Equation 3.”
Explaining Black Box Predictions and Unveiling Data Artifacts through Influence Functions (2020, background): “Beyond NLI, the negative inﬂu-ence of artifacts in data was explored in other text classiﬁcation tasks (Pryzant et al., 2018; Kumar et al., 2019; Landeiro et al., 2019), focusing on approaches to adversarial learning to demote the artifacts.”
Unsupervised Discovery of Implicit Gender Bias (2020, methodology): “We refer to Kumar et al. (2019) for the training procedure that alternates minimizing each objective.”
On Transferability of Bias Mitigation Effects in Language Model Fine-Tuning (2021, background): “…approaches reduce the inﬂuence of spurious features in the original dataset (Dixon et al., 2018; Zhao et al., 2018; Park et al., 2018), and adversarial learning approaches generate debiased data representations that are exclusive to the downstream model (Kumar et al., 2019; Zhang et al., 2018).”
Handling Bias in Toxic Speech Detection: A Survey (2022, methodology): “[116] used an adversarial training [53] on the Founta100k dataset.”
Influence Tuning: Demoting Spurious Correlations via Instance Attribution and Instance-Driven Updates (2021, background): “…(1) model-based approaches to learn confound-invariant representations, e.g., adversarial training (Pryzant et al., 2018; Elazar and Goldberg, 2018; Kumar et al., 2019); (2) data-based approaches to balance the training data, e.g., counterfactual data augmentation (Zmigrod et al., 2019; Kaushik…”