Distribution Matching for Rationalization

Huang, Yongfeng; Chen, Yujun; Du, Yulun; Yang, Zhilin

doi:10.1609/aaai.v35i14.17547

Cited by 3 publications

(8 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…with respect to the selections where the first term penalizes the number of selections, and the second one encourages continuity of selections. The specific form may be a different subject to the different architecture of the rationale model (Bastings, Aziz, and Titov 2019;Huang et al 2021;Sha, Camburu, and Lukasiewicz 2022). For example, many methods replace the first term of the above regularizer with…”

Section: Methodology Preliminarymentioning

confidence: 99%

“…Some researchers have accomplished this by adding extra modules to the rationalization process. Huang et al (2021) matched the distributions of rationales and input text in both the feature and output spaces. Sha, Camburu, and Lukasiewicz (2022) introduced an adversarial-based technique to make the select-then-predict model learn from an extra predictor.…”

Section: Selective Rationalizationmentioning

confidence: 99%

“…Datasets Following the work of Huang et al (2021) and Liu et al (2022), we consider two widely used datasets for selective rationalization. 1) BeerAdvocate (McAuley, Leskovec, and Jurafsky 2012) contains more than 220,000 beer reviews, where users rate different aspects of beer from 0 to 5 stars.…”

Section: Experiments Experimental Settingsmentioning

confidence: 99%

“…It contains reviews of hotels and ratings in different aspects. For a fair comparison, we consider both the above beer and hotel reviews prediction as a binary classification task as Bao et al (2018) did, while other settings followed the previous works (Huang et al 2021;Yu et al 2021). The Appendix has pre-processing settings details.…”

Section: Experiments Experimental Settingsmentioning

confidence: 99%

“…For example, the case of Figure 2 (Sub-optimal) given the right sentiment prediction but using the rationale missed some critical context. Many approaches (Huang et al 2021;Yu et al 2021;Yue et al 2022;Sha, Camburu, and Lukasiewicz 2022) have been proposed to address the above degeneration issue. The basic idea of these approaches is to restrain the predictor using supplementary modules that utilize full information of inputs such that the predictor does not entirely rely on rationales at training time.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Learning Robust Rationales for Model Explainability: A Guidance-Based Approach

Hu,

2024

AAAI

View full text Add to dashboard Cite

Selective rationalization can be regarded as a straightforward self-explaining approach for enhancing model explainability in natural language processing tasks. It aims to provide explanations that are more accessible and understandable to non-technical users by first selecting subsets of input texts as rationales and then predicting based on chosen subsets. However, existing methods that follow this select-then-predict framework may suffer from the rationalization degeneration problem, resulting in sub-optimal or unsatisfactory rationales that do not align with human judgments. This problem may further lead to rationalization failure, resulting in meaningless rationales that ultimately undermine people's trust in the rationalization model. To address these challenges, we propose a Guidance-based Rationalization method (G-RAT) that effectively improves robustness against failure situations and the quality of rationales by using a guidance module to regularize selections and distributions. Experimental results on two synthetic settings prove that our method is robust to the rationalization degeneration and failure problems, while the results on two real datasets show its effectiveness in providing rationales in line with human judgments. The source code is available at https://github.com/shuaibo919/g-rat.

show abstract

Section: Methodology Preliminarymentioning

confidence: 99%

Section: Selective Rationalizationmentioning

confidence: 99%

Section: Experiments Experimental Settingsmentioning

confidence: 99%

Section: Experiments Experimental Settingsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Learning Robust Rationales for Model Explainability: A Guidance-Based Approach

Hu,

2024

AAAI

View full text Add to dashboard Cite

show abstract

Decoupled Rationalization with Asymmetric Learning Rates: A Flexible Lipschitz Restraint

Liu,

Wang,

Wang

et al. 2023

Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining

View full text Add to dashboard Cite

Interventional Rationalization

Yue,

Liu,

Wang

et al. 2023

Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing

View full text Add to dashboard Cite

Selective rationalizations improve the explainability of neural networks by selecting a subsequence of the input (i.e., rationales) to explain the prediction results. Although existing methods have achieved promising results, they still suffer from adopting the spurious correlations in data (aka., shortcuts) to compose rationales and make predictions. Inspired by the causal theory, in this paper, we develop an interventional rationalization (Inter-RAT) to discover the causal rationales. Specifically, we first analyse the causalities among the input, rationales and results with a causal graph. Then, we discover spurious correlations between the input and rationales, and between rationales and results, respectively, by identifying the confounder in the causalities. Next, based on the backdoor adjustment, we propose a causal intervention method to remove the spurious correlations between input and rationales. Further, we discuss reasons why spurious correlations between the selected rationales and results exist by analysing the limitations of the sparsity constraint in the rationalization, and employ the causal intervention method to remove these correlations. Extensive experimental results on three realworld datasets clearly validate the effectiveness of our proposed method. The source code of Inter-RAT is available at https://github. com/yuelinan/Codes-of-Inter-RAT.

show abstract

Distribution Matching for Rationalization

Cited by 3 publications

References 26 publications

Learning Robust Rationales for Model Explainability: A Guidance-Based Approach

Learning Robust Rationales for Model Explainability: A Guidance-Based Approach

Decoupled Rationalization with Asymmetric Learning Rates: A Flexible Lipschitz Restraint

Interventional Rationalization

Contact Info

Product

Resources

About