Abstract
We introduce an adversarial method for producing high-recall explanations of neural text classifier decisions. Building on an existing architecture for extractive explanations via hard attention, we add an adversarial layer which scans the residual of the attention for remaining predictive signal. Motivated by the important domain of detecting personal attacks in social media comments, we additionally demonstrate the importance of manually setting a semantically appropriate “default” behavior for the model by explicitly manipulating its bias term. We develop a validation set of human-annotated personal attacks to evaluate the impact of these changes.- Anthology ID:
- D18-1386
- Volume:
- Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing
- Month:
- October-November
- Year:
- 2018
- Address:
- Brussels, Belgium
- Venue:
- EMNLP
- SIG:
- SIGDAT
- Publisher:
- Association for Computational Linguistics
- Note:
- Pages:
- 3497–3507
- Language:
- URL:
- https://aclanthology.org/D18-1386
- DOI:
- 10.18653/v1/D18-1386
- Cite (ACL):
- Samuel Carton, Qiaozhu Mei, and Paul Resnick. 2018. Extractive Adversarial Networks: High-Recall Explanations for Identifying Personal Attacks in Social Media Posts. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, pages 3497–3507, Brussels, Belgium. Association for Computational Linguistics.
- Cite (Informal):
- Extractive Adversarial Networks: High-Recall Explanations for Identifying Personal Attacks in Social Media Posts (Carton et al., EMNLP 2018)
- PDF:
- https://preview.aclanthology.org/ingestion-script-update/D18-1386.pdf