Counterfactual Inference for Text Classification Debiasing

Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, pages 5434–5445 August 1–6, 2021. ©2021 Association for Computational Linguistics 5434 Counterfactual Inference for Text Classification Debiasing Chen Qian Fuli Feng∗ Lijie Wen∗ Tsinghua University National University of Singapore Tsinghua University qc16@mails.tsinghua.edu.cn fulifeng93@gmail.com wenlj@tsinghua.edu.cn Chunping Ma Pengjun Xie Alibaba DAMO Academy Alibaba DAMO Academy chunping.mcp@alibaba-inc.com chengchen.xpjg@taobao.com Abstract Today’s text classifiers inevitably suffer from unintended dataset biases, especially the document-level label bias and word-level key- word bias, which may hurt models’ general- ization. Many previous studies employed data- level manipulations or model-level balancing mechanisms to recover unbiased distributions and thus prevent models from capturing the two types of biases. Unfortunately, they ei- ther suffer from the extra cost of data col- lection/selection/annotation or need an elab- orate design of balancing strategies. Differ- ent from traditional factual inference in which debiasing occurs before or during training, counterfactual inference mitigates the influ- ence brought by unintended confounders after training, which can make unbiased decisions with biased observations. Inspired by this, we propose a model-agnostic text classifica- tion debiasing framework – CORSAIR, which can effectively avoid employing data manip- ulations or designing balancing mechanisms. Concretely, CORSAIR first trains a base model on a training set directly, allowing the dataset biases “poison” the trained model. In infer- ence, given a factual input document, COR- SAIR imagines its two counterfactual counter- parts to distill and mitigate the two biases cap- tured by the poisonous model. Extensive ex- periments demonstrate CORSAIR’s effective- ness, generalizability and fairness. 1 1 Introduction Text classification, mapping text documents to a set of predefined categories, is a fundamental and important technique serving for many applications such as sentiment analysis (Qian et al., 2020b), ∗This work was partly done during Chen Qian’s intern- ship at Alibaba DAMO academy. Fuli Feng and Lijie Wen are the co-corresponding authors. 1The code is available at https://github.com/ qianc62/Corsair. partisanship recognition (Kiesel et al., 2019) and spam detection (Castillo et al., 2007). Machine learning models have become the default choice of solving text classification, owing to their abil- ity to recognize the textual patterns from the la- beled documents (Kim, 2014; Howard and Ruder, 2018). Nevertheless, they are at the risk of inad- vertently capturing and even amplifying the unin- tended dataset biases (Zhao et al., 2017; Zhang et al., 2020; Feder et al., 2020; Blodgett et al., 2020), which can be at document-level (i.e., label bias) and word-level (i.e., keyword bias). The label bias issue occurs in the scenarios where a portion of the categories possesses a ma- jority of training examples than others. For ex- ample, the label distribution of a binary sentiment analysis dataset could be 95%:5% (Dixon et al., 2018). Many previous studies found that the mod- els trained on such data are potentially at the risk of simply predicting the majority answers (Dixon et al., 2018; Zhang et al., 2020). The keyword bias issue occurs in the situation where trained models …

Content truncated to 500 words