--- title: "Towards Robust and Privacy-preserving Text — Li 2018" description: "Written text reveals details about its author, such as gender, age and location." authors: ["Yitong Li", "Timothy Baldwin", "Trevor Cohn"] year: 2018 venue: "Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers)" source: "https://aclanthology.org/P18-2005/" doi: "10.18653/v1/p18-2005" licence: "CC BY 4.0 (ACL Anthology)" licence_url: "https://creativecommons.org/licenses/by/4.0/" summary_licence: "Elamur Cosmic License" canonical: "https://research.elamur.ai/free/en/papers/li-2018-robust-privacy-preserving-text-representations/" --- [Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/li-2018-robust-privacy-preserving-text-representations/) [home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2018](https://md.research.elamur.ai/free/en/papers/2018/page/4/) / Towards Robust and Privacy-preserving Text Representations # Towards Robust and Privacy-preserving Text Representations **Authors:** Yitong Li, [Timothy Baldwin](https://md.research.elamur.ai/free/en/authors/timothy-baldwin/), [Trevor Cohn](https://md.research.elamur.ai/free/en/authors/trevor-cohn/) · **Year:** 2018 · **Venue:** Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/) **Links:** [Publisher's page](https://aclanthology.org/P18-2005/) · [PDF](https://aclanthology.org/P18-2005.pdf) · [DOI 10.18653/v1/p18-2005](https://doi.org/10.18653/v1/p18-2005) · [OpenAlex](https://openalex.org/W2963879260) **Citations:** 152 citations (openalex-snapshot · 2026-10-14) · [81](#cited-by) in our library, of the 140884 papers whose bibliographies we read ## Summary Written text reveals details about its author, such as gender, age and location. NLP models can use these traits as shortcuts. This hurts performance on texts by unseen authors or from new domains, and it lets sensitive information leak through shared hidden representations. Li, Baldwin and Cohn address both problems with adversarial training based on Ganin et al. (2016). Each protected attribute gets its own discriminator, which tries to predict that attribute from the representation. A gradient reversal layer then pushes the encoder to defeat these discriminators while still serving the main task. The method is tested on two tasks: - **POS tagging:** a bi-LSTM pre-trained on the Web English Treebank, trained on TrustPilot data, and also evaluated on a held-out AAVE corpus. - **Sentiment:** 5-class rating prediction on TrustPilot reviews with a Kim (2014) CNN. Adversarial training narrows tagging accuracy gaps between demographic groups and improves accuracy both in-domain and out-of-domain. It also makes private attributes harder to recover, while sentiment F1 stays essentially unchanged. Unlike differential privacy, the approach gives no formal guarantees, but it also costs no accuracy. ## Key points - **Objective:** a min–max loss with one cross-entropy adversarial term per protected attribute (λ = 10⁻³), implemented via gradient reversal. The code is released on GitHub (lrank/Robust_and_Privacy_preserving_Text_Representations). - **TrustPilot POS tagging (600 sentences, Google Universal POS tagset):** - The baseline shows a 1.5-point accuracy gap between AGE groups (O45 vs U35); ADV cuts it to 0.3. - The SEX gap falls from 0.2 to 0.1. - Group accuracies rise to about 92%. - **AAVE held-out evaluation:** ADV improves accuracy on every domain: - LYRICS: 73.7 → 80.5 - SUBTITLES: 81.4 → 85.8 - TWEETS: 59.9 → 65.4 - Macro-average: 71.7 → 77.0 - **Sentiment data:** TrustPilot reviews filtered to English with LANGID.PY (confidence > 0.9) and balanced to 10k reviews for each of 5 countries. - **Sentiment results:** - Location attack accuracy drops from 53.4% to 22.1% with ADV-LOC, close to the 20% majority baseline. - Test F1 stays around 40% (40.1 baseline vs 40.2 for ADV-all). *Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.* Fields: [Natural language processing](https://md.research.elamur.ai/free/en/shelves/nlp/) Topics: [Part-of-speech tagging](https://md.research.elamur.ai/free/en/topics/part-of-speech-tagging/) ## Cited by Cited by 81 papers of our library cite it (of the 140884 whose bibliographies we read); 8 of them have a page here, 73 not yet. - [Don't Take the Premise for Granted: Mitigating Artifacts in Natural Language Inference](https://md.research.elamur.ai/free/en/papers/belinkov-2019-don-t-take-premise-granted/) - [Adversarial Removal of Demographic Attributes from Text Data](https://md.research.elamur.ai/free/en/papers/elazar-2018-adversarial-removal-demographic-attributes-text/) - [The Importance of Modeling Social Factors of Language: Theory and Practice](https://md.research.elamur.ai/free/en/papers/hovy-2021-importance-modeling-social-factors-language/) - [Deduplicating Training Data Mitigates Privacy Risks in Language Models](https://md.research.elamur.ai/free/en/papers/kandpal-2022-deduplicating-training-data-mitigates-privacy/) - [Debiasing Pre-trained Contextualised Embeddings](https://md.research.elamur.ai/free/en/papers/kaneko-2021-debiasing-pre-trained-contextualised-embeddings/) - [Gender Bias in Machine Translation](https://md.research.elamur.ai/free/en/papers/savoldi-2021-gender-bias-machine-translation/) - [Predictive Biases in Natural Language Processing Models: A Conceptual Framework and Overview](https://md.research.elamur.ai/free/en/papers/shah-2019-predictive-biases-natural-language-processing/) - [Mitigating Gender Bias in Natural Language Processing: Literature Review](https://md.research.elamur.ai/free/en/papers/sun-2019-mitigating-gender-bias-natural-language/) ## Cite The paper itself, not this page. BibTeX from the library's record of the work. ```bibtex @inproceedings{li2018robust, title = {{Towards Robust and Privacy-preserving Text Representations}}, author = {Li, Yitong and Baldwin, Timothy and Cohn, Trevor}, booktitle = {Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers)}, year = {2018}, doi = {10.18653/v1/p18-2005}, url = {https://doi.org/10.18653/v1/p18-2005} } ``` - [← previous Towards JointUD: Part-of-speech Tagging and Lemmatization using Recurrent Neural Networks](https://md.research.elamur.ai/free/en/papers/ud-2018-arakelyan-jointud-part-speech-tagging-lemmatization/) - [next → Training Millions of Personalized Dialogue Agents](https://md.research.elamur.ai/free/en/papers/mazare-2018-training-millions-personalized-dialogue-agents/) - [↑ back to papers of 2018](https://md.research.elamur.ai/free/en/papers/2018/page/4/) ## References 2 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet. 1. users based on their hidden representation. That is, lower discriminator performance indicates that the representation conveys better privacy for individuals, and vice versa. [find](https://scholar.google.com/scholar?q=users+based+on+their+hidden+representation%2E+That+is%2C+lower+discriminator+performance+indicates+that+the+representation+conveys+better+privacy+for+individuals%2C+and+vice+versa%2E) 2. Martin Abadi, Andy Chu, Ian Goodfellow, H Brendan McMahan, Ilya Mironov, Kunal Talwar, and Li Zhang. 2016. Deep learning with differential privacy. In Proceedings of the 2016 ACM SIGSAC Conference on Computer and Communications Security, pages 308–318. [find](https://scholar.google.com/scholar?q=Martin+Abadi%2C+Andy+Chu%2C+Ian+Goodfellow%2C+H+Brendan+McMahan%2C+Ilya+Mironov%2C+Kunal+Talwar%2C+and+Li+Zhang%2E+2016%2E+Deep+learning+with+differential+privacy%2E+In+Proceedings+of+the+2016+ACM+SIGSAC+Conference+on+)