广西科学院学报2025，Vol.41Issue(4)：363-375,13.DOI:10.13657/j.cnki.gxkxyxb.20260107.001
基于因果分析的文本去偏技术研究综述
A Review of Text Debiasing Technologies Based on Causal Analy-sis
摘要
Abstract
Deep learning model in Natural Language Processing(NLP)task is prone to misidentifying surface-level correlations as causal relationships.This leads to the continuous accumulation of biases derived from linguistic patterns,label co-occurrences,and corpus distribution,which ultimately undermines the mod-els' generalization,fairness,and interpretability.Consequently,there is an urgent need for systematic de-bia-sing mechanisms to eliminate these biases.The text debiasing technology based on causal analysis have grad-ually developed in this context.This paper systematically reviews the development process of debiasing tech-nology from empirical paradigms such as data augmentation and regularization to causal graph-driven para-digm.Through the"causal graph modeling—effect estimation—causal intervention"method for text tasks,the bias problem in text tasks is systematically analyzed and dealt with.On this basis,we focus on three ma-instream technical paths of counterfactual debiasing,back-door adjustment and front-door adjustment.At the task level,text classification,sentiment analysis and fact verification are selected as representative scenarios,and the typical debiasing methods of the three technical paths are correspondingly discussed.The typical methods are compared and analyzed from bias types,debiasing methods,advantages and limitations of core in-tervention strategies.Based on the existing research,the author believes that the current causal text debiasing technology still has the following problems.There is still a lack of multi-source biases collaborative model-ing.The generation of counterfactual samples is difficult to strike a balance between semantic preservation and generation cost.The causal structure relies too much on expert priors.Scalability is limited in multi-hop reasoning,cross-lingual and multimodal scenarios.In view of the above shortcomings,this article proposes some improvement measures from the aspects of unified multi-source causal modeling,high-quality counter-factual generation with semantic preservation,automated causal structure learning with robust effect estima-tion as well as lightweight causal debiasing mechanism for large-scale models and large-scale applications.The research prospect of deep integration of causal reasoning with large language model and multimodal model is prospected.关键词
因果推断/自然语言处理(NLP)/反事实推理/后门调整/前门调整/文本去偏/模型公平性Key words
causal inference/Natural Language Processing(NLP)/counterfactual inference/back-door adjust-ment/front-door adjustment/text debiasing/model fairness分类
信息技术与安全科学引用本文复制引用
元昌安,赵剑波,蔡宏果,彭昱忠..基于因果分析的文本去偏技术研究综述[J].广西科学院学报,2025,41(4):363-375,13.基金项目
国家自然科学基金项目(62262044)和广西自然科学基金项目(2023GXNSFAA026027)资助. （62262044）
