https://github.com/deep-spin/crest

ACL 2023

CREST:
A Joint Framework for Rationalization
and Counterfactual Text Generation

Nuno M. Guerreiro

André F. T. Martins

Alexis Ross

Marcos Treviso

How to explain the classifier's decision?

This album is terrible and some of the songs are really bad

NEG

NEU

POS

How to explain the classifier's decision?

This album is terrible and some of the songs are really bad
This album is terrible and some of the songs are really bad

NEG

NEU

POS

​Selective Rationalization

a.k.a. mask-then-predict

highlights as explanations

How to explain the classifier's decision?

This album is terrible and some of the songs are really bad
This album is terrible and some of the songs are really bad 
This album is amazing and some of the songs are very well written

​Selective Rationalization

a.k.a. mask-then-predict

highlights as explanations

Counterfactual Generation

a.k.a. what-if questions

edits as explanations

NEG

NEU

POS

How to explain the classifier's decision?

CREST (ContRastive Edits with Sparse raTionalization)

​Selective Rationalization

a.k.a. mask-then-predict

highlights as explanations

Counterfactual Generation

a.k.a. what-if questions

edits as explanations

NEG

NEU

POS

This album is terrible and some of the songs are really bad
This album is terrible and some of the songs are really bad 
This album is amazing and some of the songs are very well written

CREST-Generation

This album is terrible and some of the songs are really bad

\(\bm{x}\)

* 2-stage process inspired by MiCE:
Ross, A., Marasović, A., & Peters, M. E. Explaining NLP models via minimal contrastive editing (MiCE). Findings of ACL 2021

CREST-Generation

Trainable Masker

This album is terrible and some of the songs are really bad

\(\bm{x}\)

* 2-stage process inspired by MiCE:
Ross, A., Marasović, A., & Peters, M. E. Explaining NLP models via minimal contrastive editing (MiCE). Findings of ACL 2021

CREST-Generation

Trainable Masker

\(\bm{z}\)

This album is terrible and some of the songs are really bad

\(\bm{x}\)

This album is terrible and some of the songs are really bad

* 2-stage process inspired by MiCE:
Ross, A., Marasović, A., & Peters, M. E. Explaining NLP models via minimal contrastive editing (MiCE). Findings of ACL 2021

CREST-Generation

\(\hat{y}\)

NEG

Predictor

Trainable Masker

\(\bm{z}\)

This album is terrible and some of the songs are really bad

\(\bm{x}\)

This album is terrible and some of the songs are really bad

* 2-stage process inspired by MiCE:
Ross, A., Marasović, A., & Peters, M. E. Explaining NLP models via minimal contrastive editing (MiCE). Findings of ACL 2021

CREST-Generation

\(\hat{y}\)

NEG

Predictor

Trainable Masker

\(\bm{z}\)

This album is terrible and some of the songs are really bad

\(\bm{x}\)

POS: This album is MASK 
and some of the songs 
are MASK
This album is terrible and some of the songs are really bad

* 2-stage process inspired by MiCE:
Ross, A., Marasović, A., & Peters, M. E. Explaining NLP models via minimal contrastive editing (MiCE). Findings of ACL 2021

CREST-Generation

\(\hat{y}\)

NEG

Predictor

Trainable Masker

\(\bm{z}\)

This album is terrible and some of the songs are really bad

\(\bm{x}\)

\(\tilde{\bm{x}}\)

POS: This album is MASK 
and some of the songs 
are MASK

Editor

This album is terrible and some of the songs are really bad 
This album is amazing and some of the songs are very well written

* 2-stage process inspired by MiCE:
Ross, A., Marasović, A., & Peters, M. E. Explaining NLP models via minimal contrastive editing (MiCE). Findings of ACL 2021

Experiments

• Masker: SPECTRA rationalizer [1]

– Controllable sparsity

– Contiguity

– Robustness

[1] Guerreiro, N.M. and Martins, A. F. T. SPECTRA: Sparse structured text rationalization. EMNLP 2021.

Experiments

• Masker: SPECTRA rationalizer [1]
• Editor: Pre-trained T5 model

– Controllable sparsity

– Contiguity

– Robustness

[1] Guerreiro, N.M. and Martins, A. F. T. SPECTRA: Sparse structured text rationalization. EMNLP 2021.

Experiments

• Masker: SPECTRA rationalizer [1]
• Editor: Pre-trained T5 model

• Two datasets: IMDB and SNLI

– Controllable sparsity

– Contiguity

– Robustness

[1] Guerreiro, N.M. and Martins, A. F. T. SPECTRA: Sparse structured text rationalization. EMNLP 2021.

Experiments

• Masker: SPECTRA rationalizer [1]
• Editor: Pre-trained T5 model

• Two datasets: IMDB and SNLI

– Controllable sparsity

– Contiguity

– Robustness

[1] Guerreiro, N.M. and Martins, A. F. T. SPECTRA: Sparse structured text rationalization. EMNLP 2021.

Experiments

• Masker: SPECTRA rationalizer [1]
• Editor: Pre-trained T5 model

• Two datasets: IMDB and SNLI

– Controllable sparsity

– Contiguity

– Robustness

[1] Guerreiro, N.M. and Martins, A. F. T. SPECTRA: Sparse structured text rationalization. EMNLP 2021.

Experiments

• Masker: SPECTRA rationalizer [1]
• Editor: Pre-trained T5 model

• Two datasets: IMDB and SNLI

– Controllable sparsity

– Contiguity

– Robustness

[1] Guerreiro, N.M. and Martins, A. F. T. SPECTRA: Sparse structured text rationalization. EMNLP 2021.

Experiments

• Masker: SPECTRA rationalizer [1]
• Editor: Pre-trained T5 model

• Two datasets: IMDB and SNLI

– Controllable sparsity

– Contiguity

– Robustness

[1] Guerreiro, N.M. and Martins, A. F. T. SPECTRA: Sparse structured text rationalization. EMNLP 2021.

Human Evaluation

• 100 examples (IMDB + SNLI)

• Validity and naturalness (e.g., based on style, tone, and grammar)

• 5-point Likert scale

Human Evaluation

• 100 examples (IMDB + SNLI)

• Validity and naturalness (e.g., based on style, tone, and grammar)

• 5-point Likert scale

Human Evaluation

• 100 examples (IMDB + SNLI)

• Validity and naturalness (e.g., based on style, tone, and grammar)

• 5-point Likert scale

How to leverage counterfactuals?

CREST-Generation

✅

High-quality counterfactuals

How to leverage counterfactuals?

CREST-Generation

???

✅

High-quality counterfactuals

How to leverage counterfactuals?

CREST-Generation

🤔

Data augmentation?

✅

High-quality counterfactuals

How to leverage counterfactuals?

CREST-Generation

✅

High-quality counterfactuals

🥰

Exploit the paired structure of factual and counterfactual inputs

CREST-Rationalization

CREST-Rationalization

CREST-Generation

CREST-Rationalization

\(\bm{z}^\star\)

\(\bm{x}\)

This album is terrible and some of the songs are really bad

CREST-Generation

CREST-Rationalization

\(\bm{z}^\star\)

\(\bm{x}\)

This album is terrible and some of the songs are really bad 
This album is amazing and some of the songs are very well written

CREST-Generation

\(\tilde{\bm{x}}\)

\(\tilde{\bm{z}}^\star\)

CREST-Rationalization

\(\bm{z}^\star\)

\(\bm{x}\)

This album is terrible and some of the songs are really bad 
This album is amazing and some of the songs are very well written

CREST-Generation

\(\tilde{\bm{x}}\)

\(\tilde{\bm{z}}^\star\)

FACTUAL  
FLOW

COUNTERFACTUAL  
FLOW

CREST-Rationalization

\(\bm{z}^\star\)

\(\bm{x}\)

This album is terrible and some of the songs are really bad 
This album is amazing and some of the songs are very well written

CREST-Generation

\(\tilde{\bm{x}}\)

\(\tilde{\bm{z}}^\star\)

FACTUAL  
FLOW

COUNTERFACTUAL  
FLOW

Trainable Masker

Trainable Masker

shared

CREST-Rationalization

\(\bm{z}^\star\)

\(\bm{x}\)

This album is terrible and some of the songs are really bad 
This album is amazing and some of the songs are very well written

CREST-Generation

\(\tilde{\bm{x}}\)

\(\tilde{\bm{z}}^\star\)

FACTUAL  
FLOW

COUNTERFACTUAL  
FLOW

Trainable Masker

Trainable Masker

\(\bm{z}\)

\(\tilde{\bm{z}}\)

This album is terrible and some of the songs are really bad 
This album is amazing and some of the songs are very well written

shared

CREST-Rationalization

\(\bm{z}^\star\)

\(\bm{x}\)

This album is terrible and some of the songs are really bad 
This album is amazing and some of the songs are very well written

CREST-Generation

\(\tilde{\bm{x}}\)

\(\tilde{\bm{z}}^\star\)

FACTUAL  
FLOW

COUNTERFACTUAL  
FLOW

Trainable Masker

Trainable Masker

\(\bm{z}\)

\(\tilde{\bm{z}}\)

This album is terrible and some of the songs are really bad 
This album is amazing and some of the songs are very well written
NEG

Predictor

POS

Predictor

\(\hat{y}\)

\(\tilde{y}\)

shared

shared

CREST-Rationalization

\(\bm{z}^\star\)

\(\bm{x}\)

This album is terrible and some of the songs are really bad 
This album is amazing and some of the songs are very well written

CREST-Generation

\(\tilde{\bm{x}}\)

\(\tilde{\bm{z}}^\star\)

FACTUAL  
FLOW

COUNTERFACTUAL  
FLOW

Trainable Masker

Trainable Masker

\(\bm{z}\)

\(\tilde{\bm{z}}\)

This album is terrible and some of the songs are really bad 
This album is amazing and some of the songs are very well written
NEG

Predictor

POS

Predictor

\(\hat{y}\)

\(\tilde{y}\)

shared

shared

\Omega(\bm{z}, \tilde{\bm{z}}) = \|\bm{z} - \bm{z}^\star \|^2 + \|\tilde{\bm{z}} - \tilde{\bm{z}}^\star \|^2

Agreement regularization:

factual similarity

counterfactual similarity

Experiments

• SPECTRA rationalizer as the masker
• Experiments on IMDB and SNLI

Original inputs

Original + Human CFs

Original + CREST CFs

CREST-Rationalization

Data Augmentation

in-domain

IMDB

Revised IMDB

Contrast IMDB

RotTom

SST-2

Amazon

Yelp

contrastive

out-of-domain

Experiments

• SPECTRA rationalizer as the masker
• Experiments on IMDB and SNLI

Original inputs

Original + Human CFs

Original + CREST CFs

CREST-Rationalization

Data Augmentation

in-domain

IMDB

Revised IMDB

Contrast IMDB

RotTom

SST-2

Amazon

Yelp

contrastive

out-of-domain

Experiments

• SPECTRA rationalizer as the masker
• Experiments on IMDB and SNLI

Original inputs

Original + Human CFs

Original + CREST CFs

CREST-Rationalization

Data Augmentation

in-domain

IMDB

Revised IMDB

Contrast IMDB

RotTom

SST-2

Amazon

Yelp

contrastive

out-of-domain

Experiments

• SPECTRA rationalizer as the masker
• Experiments on IMDB and SNLI

Original inputs

Original + Human CFs

Original + CREST CFs

CREST-Rationalization

Data Augmentation

in-domain

IMDB

Revised IMDB

Contrast IMDB

RotTom

SST-2

Amazon

Yelp

contrastive

out-of-domain

Experiments

• SPECTRA rationalizer as the masker
• Experiments on IMDB and SNLI

Original inputs

Original + Human CFs

Original + CREST CFs

CREST-Rationalization

Data Augmentation

in-domain

IMDB

Revised IMDB

Contrast IMDB

RotTom

SST-2

Amazon

Yelp

contrastive

out-of-domain

Experiments

• SPECTRA rationalizer as the masker
• Experiments on IMDB and SNLI

Original inputs

Original + Human CFs

Original + CREST CFs

CREST-Rationalization

Data Augmentation

in-domain

IMDB

Revised IMDB

Contrast IMDB

RotTom

SST-2

Amazon

Yelp

contrastive

out-of-domain

Experiments

• SPECTRA rationalizer as the masker
• Experiments on IMDB and SNLI

Original inputs

Original + Human CFs

Original + CREST CFs

CREST-Rationalization

Data Augmentation

in-domain

IMDB

Revised IMDB

Contrast IMDB

RotTom

SST-2

Amazon

Yelp

contrastive

out-of-domain

Interpretability Analysis

Are the rationales generated by CREST-Rationalization interpretable?

Interpretability Analysis

Are the rationales generated by CREST-Rationalization interpretable?

• Plausibility (human-likeness)


• Forward simulability (informativeness)

• Counterfactual simulability:

Interpretability Analysis

Are the rationales generated by CREST-Rationalization interpretable?

\dfrac{1}{N} \sum\limits_{i=1}^N [[ C(\bm{x}_i) \neq C(G(\bm{x}_i \odot \bm{z}_i)) ]]

input with infilled tokens by \(G\) in the positions marked by \(\bm{z}\)

• Plausibility (human-likeness)


• Forward simulability (informativeness)

• Counterfactual simulability:

ability of a rationale \(\bm{z}\) to change the classifier's prediction when the classifier receives a rationale-guided contrastive edit as input

Interpretability Analysis

Are the rationales generated by CREST-Rationalization interpretable?

\(F\)

\(F + C_H\)

\(F + C_S\)

\(F \,\&\, C_S\)

0.6733 ± 0.02                        91.70 ± 0.92                          81.18 ± 2.79

0.6718 ± 0.04                        91.44 ± 1.46                          80.53 ± 4.17

0.6758 ± 0.01                        91.68 ± 0.59                          84.54 ± 1.09

0.6904 ± 0.02                        91.93 ± 0.83                          86.43 ± 1.56

Setup

Plausibility
(AUC)

Forward Sim.
(ACC)

Counterfactual Sim.
(ACC)

Data augmentation:

CREST-Rationalization:

Interpretability Analysis

Are the rationales generated by CREST-Rationalization interpretable?

\(F\)

\(F + C_H\)

\(F + C_S\)

\(F \,\&\, C_S\)

0.6733 ± 0.02                        91.70 ± 0.92                          81.18 ± 2.79

0.6718 ± 0.04                        91.44 ± 1.46                          80.53 ± 4.17

0.6758 ± 0.01                        91.68 ± 0.59                          84.54 ± 1.09

0.6904 ± 0.02                        91.93 ± 0.83                          86.43 ± 1.56

Setup

Plausibility
(AUC)

Forward Sim.
(ACC)

Counterfactual Sim.
(ACC)

Data augmentation:

CREST-Rationalization:

Interpretability Analysis

Are the rationales generated by CREST-Rationalization interpretable?

\(F\)

\(F + C_H\)

\(F + C_S\)

\(F \,\&\, C_S\)

0.6733 ± 0.02                        91.70 ± 0.92                          81.18 ± 2.79

0.6718 ± 0.04                        91.44 ± 1.46                          80.53 ± 4.17

0.6758 ± 0.01                        91.68 ± 0.59                          84.54 ± 1.09

0.6904 ± 0.02                        91.93 ± 0.83                          86.43 ± 1.56

Setup

Plausibility
(AUC)

Forward Sim.
(ACC)

Counterfactual Sim.
(ACC)

Data augmentation:

CREST-Rationalization:

Conclusions

     •  produces valid, fluent, and diverse counterfactuals
     •  controls the amount of perturbation
     •  leads to plausible explanations
     •  achieves high counterfactual simulability

https://arxiv.org/abs/2305.17075

https://github.com/deep-spin/crest

Selective Rationalization ❤︎ Counterfactual Generation

Experiments

• Masker: SPECTRA rationalizer [1]
• Editor: Pre-trained T5 model

[1] Guerreiro, N.M. and Martins, A. F. T. SPECTRA: Sparse structured text rationalization. EMNLP 2021.

– Controllable sparsity

– Contiguity

– Robustness

Experiments

• Masker: SPECTRA rationalizer [1]
• Editor: Pre-trained T5 model

[1] Guerreiro, N.M. and Martins, A. F. T. SPECTRA: Sparse structured text rationalization. EMNLP 2021.

– Controllable sparsity

– Contiguity

– Robustness

validity

fluency

diversity

closeness

Experiments

• Masker: SPECTRA rationalizer [1]
• Editor: Pre-trained T5 model

[1] Guerreiro, N.M. and Martins, A. F. T. SPECTRA: Sparse structured text rationalization. EMNLP 2021.

– Controllable sparsity

– Contiguity

– Robustness

validity

fluency

diversity

closeness

Experiments

• Masker: SPECTRA rationalizer [1]
• Editor: Pre-trained T5 model

[1] Guerreiro, N.M. and Martins, A. F. T. SPECTRA: Sparse structured text rationalization. EMNLP 2021.

– Controllable sparsity

– Contiguity

– Robustness

validity

fluency

diversity

closeness

Experiments

• Masker: SPECTRA rationalizer [1]
• Editor: Pre-trained T5 model

[1] Guerreiro, N.M. and Martins, A. F. T. SPECTRA: Sparse structured text rationalization. EMNLP 2021.

– Controllable sparsity

– Contiguity

– Robustness

validity

fluency

diversity

closeness

Experiments

• SPECTRA rationalizer as the masker
• Experiments on IMDB and SNLI

in-domain    contrast    out-of-domain

Experiments

• SPECTRA rationalizer as the masker
• Experiments on IMDB and SNLI

in-domain    contrast    out-of-domain

\(F\)

\(F + C_H\)

\(F + C_S\)

\(F \,\&\, C_S\)

Data augmentation:

CREST-Rationalization:

Setup

IMDB

Revised IMDB

Contrast IMDB

RotTom

SST-2

Amazon

Yelp

Experiments

• SPECTRA rationalizer as the masker
• Experiments on IMDB and SNLI

in-domain    contrast    out-of-domain

\(F\)

\(F + C_H\)

\(F + C_S\)

\(F \,\&\, C_S\)

Data augmentation:

CREST-Rationalization:

Setup

IMDB

Revised IMDB

Contrast IMDB

RotTom

SST-2

Amazon

Yelp

Experiments

• SPECTRA rationalizer as the masker
• Experiments on IMDB and SNLI

in-domain    contrast    out-of-domain

\(F\)

\(F + C_H\)

\(F + C_S\)

\(F \,\&\, C_S\)

Data augmentation:

CREST-Rationalization:

Setup

IMDB

Revised IMDB

Contrast IMDB

RotTom

SST-2

Amazon

Yelp

Experiments

• SPECTRA rationalizer as the masker
• Experiments on IMDB and SNLI

in-domain    contrast    out-of-domain

\(F\)

\(F + C_H\)

\(F + C_S\)

\(F \,\&\, C_S\)

Data augmentation:

CREST-Rationalization:

Setup

IMDB

Revised IMDB

Contrast IMDB

RotTom

SST-2

Amazon

Yelp

Experiments

• SPECTRA rationalizer as the masker
• Experiments on IMDB and SNLI

in-domain    contrast    out-of-domain

\(F\)

\(F + C_H\)

\(F + C_S\)

\(F \,\&\, C_S\)

Data augmentation:

CREST-Rationalization:

Setup

IMDB

Revised IMDB

Contrast IMDB

RotTom

SST-2

Amazon

Yelp

Experiments

• SPECTRA rationalizer as the masker
• Experiments on IMDB and SNLI

in-domain    contrast    out-of-domain

\(F\)

\(F + C_H\)

\(F + C_S\)

\(F \,\&\, C_S\)

91.1 ± 0.3        91.4 ± 0.8       88.5 ± 0.9       76.5 ± 1.6       79.8 ± 1.6       86.0 ± 0.7       88.5 ± 0.7

90.9 ± 0.5        92.9 ± 0.9       90.4 ± 1.6       76.6 ± 1.5       80.7 ± 1.3       86.3 ± 1.0       89.1 ± 1.2

90.8 ± 0.2        91.6 ± 1.3       89.2 ± 0.4       76.7 ± 1.0       80.6 ± 0.6       86.4 ± 0.6       89.1 ± 0.5

91.2 ± 0.5        92.9 ± 0.5       89.7 ± 1.1       77.3 ± 2.3       81.1 ± 2.4       86.8 ± 0.8       89.3 ± 0.7

Data augmentation:

CREST-Rationalization:

Setup

IMDB

Revised IMDB

Contrast IMDB

RotTom

SST-2

Amazon

Yelp

Experiments

• SPECTRA rationalizer as the masker
• Experiments on IMDB and SNLI

in-domain    contrast    out-of-domain

\(F\)

\(F + C_H\)

\(F + C_S\)

\(F \,\&\, C_S\)

91.1 ± 0.3        91.4 ± 0.8       88.5 ± 0.9       76.5 ± 1.6       79.8 ± 1.6       86.0 ± 0.7       88.5 ± 0.7

90.9 ± 0.5        92.9 ± 0.9       90.4 ± 1.6       76.6 ± 1.5       80.7 ± 1.3       86.3 ± 1.0       89.1 ± 1.2

90.8 ± 0.2        91.6 ± 1.3       89.2 ± 0.4       76.7 ± 1.0       80.6 ± 0.6       86.4 ± 0.6       89.1 ± 0.5

91.2 ± 0.5        92.9 ± 0.5       89.7 ± 1.1       77.3 ± 2.3       81.1 ± 2.4       86.8 ± 0.8       89.3 ± 0.7

Data augmentation:

CREST-Rationalization:

Setup

IMDB

Revised IMDB

Contrast IMDB

RotTom

SST-2

Amazon

Yelp