Papers by Ivan Habernal
DP-BART for Privatized Text Rewriting under Local Differential Privacy (2023.findings-acl)
Copied to clipboard
| Challenge: | Existing systems that guarantee privacy protection to individuals face several problems, such as formal mathematical flaws, unrealistic privacy guarantees, and privatization of only individual words. |
| Approach: | They propose a system that enables sharing of sensitive textual documents while formally guaranteeing privacy protection to individuals. |
| Outcome: | The proposed system outperforms existing LDP systems in terms of noise and privacy guarantees. |
Granularity is crucial when applying differential privacy to text: An investigation for neural machine translation (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Differential privacy (DP) is a method that is increasingly popular in NLP but the choice of granularity at which it is applied is often neglected. |
| Approach: | They propose to apply differential privacy (DP) to a neural machine translation dataset to protect individual data points. |
| Outcome: | The proposed method is more resistant to membership inference attacks than the document-level NMT system. |
The Argument Reasoning Comprehension Task: Identification and Reconstruction of Implicit Warrants (N18-1)
Copied to clipboard
| Challenge: | Existing methods for analyzing warrants in natural language arguments are insufficient. |
| Approach: | They propose a method for reconstructing warrants from news comments . they use a crowdsourcing process to obtain warrants for 2k authentic arguments . |
| Outcome: | The proposed method will define a substantial step towards automatic warrant reconstruction. |
Private Language Models via Truncated Laplacian Mechanism (2024.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods for word embedding are prone to privacy leakage, resulting in weaker relaxations of DP that are inferior to the canonical DP in terms of privacy strength. |
| Approach: | They propose a method for private word embedding that uses a non-trivial extension of the truncated Laplacian mechanism and propose to test its effectiveness. |
| Outcome: | The proposed method has lower variance compared to the previous methods. |
Privacy-Preserving Graph Convolutional Networks for Text Classification (2022.lrec-1)
Copied to clipboard
| Challenge: | Graph convolutional networks (GCNs) are powerful for representation learning on documents that naturally occur as graphs, but sensitive personal information is prone to privacy leaks. |
| Approach: | They propose a method that adapts differentially-private gradient-based training to GCNs and conduct experiments using two optimizers on five NLP datasets in two languages. |
| Outcome: | The proposed method improves baseline privacy bounds by 2.7 while retaining competitive F1 scores while providing strong privacy guarantees. |
Before Name-Calling: Dynamics and Triggers of Ad Hominem Fallacies in Web Argumentation (N18-1)
Copied to clipboard
| Challenge: | Existing research lacks solid empirical investigation of typology of ad hominem arguments and their potential causes. |
| Approach: | They propose to perform several large-scale annotation studies and experiment with various neural architectures to validate hypotheses such as controversy or reasonableness. |
| Outcome: | The proposed model identifies the ad hominem fallacy and its possible causes using explainable neural network architectures. |
Why do you think that? Exploring Faithful Sentence-Level Rationales Without Supervision (2020.findings-emnlp)
Copied to clipboard
| Challenge: | Large pre-trained language models, such as BERT or RoBERTa, gain impressive results on a large variety of NLP tasks, including reasoning and inference. |
| Approach: | They propose a differentiable training framework to create models which output faithful rationales on a sentence level, by solely applying supervision on the target task. |
| Outcome: | The proposed model outperforms pipeline approaches and non-differentiable models on three different datasets while exceeding pipeline counterparts. |
Adapting Serious Game for Fallacious Argumentation to German: Pitfalls, Insights, and Best Practices (L18-1)
Copied to clipboard
| Challenge: | 'homeschooling' and 'death penalty' are non-existent in Germany, while being highly controversial topics of discussion in the United States. |
| Approach: | They propose to port Argotario (serious game for learning argumentation fallacies) to another language and analyze users' behavior and in-game created data to assess dissemination strategies and qualitative aspects of the resulting corpus. |
| Outcome: | The proposed game is based on a German-based game platform that can be used to learn argumentation fallacies. |
MONETA: Multimodal Industry Classification through Geographic Information with Multi Agent Systems (2026.acl-long)
Copied to clipboard
| Challenge: | MONETA is the first multimodal industry classification benchmark with text and geospatial sources. |
| Approach: | They propose a multimodal industry classification benchmark using text and geospatial sources. |
| Outcome: | The proposed model increases the accuracy of the existing models by 22.80%. |
Trade-Offs Between Fairness and Privacy in Language Modeling (2023.findings-acl)
Copied to clipboard
| Challenge: | Existing research suggests that privacy preservation comes at the price of worsening biases in classification tasks. |
| Approach: | They propose to incorporate privacy preservation and de-biasing techniques into training text generation models to investigate the trade-off between the two dimensions. |
| Outcome: | The proposed model improves on bias detection, privacy attacks, language modeling, and performance on downstream tasks. |
How reparametrization trick broke differentially-private text representation learning (2022.acl-short)
Copied to clipboard
| Challenge: | Differential privacy (DP) is a formal mathematical treatment of privacy protection . it guarantees how much privacy can be lost in the worst case . adapting DP mechanisms to NLP properly is largely non-trivial task . |
| Approach: | They propose to use differential privacy to learn text representations using DPText to quantify and guarantee how much privacy can be lost in the worst case. |
| Outcome: | The proposed methods are falsely claimed to be differentially private and violate privacy loss guarantees. |
DP-NMT: Scalable Differentially Private Machine Translation (2024.eacl-demo)
Copied to clipboard
| Challenge: | Neural machine translation (NMT) is a popular text generation task, yet there is nagging data privacy concerns. |
| Approach: | They propose an open-source framework for a privacy-preserving NMT with DP-SGD. |
| Outcome: | The proposed framework is open-source and open to the public . it combines models, datasets, and evaluation metrics to demonstrate its effectiveness. |
DP-Rewrite: Towards Reproducibility and Transparency in Differentially Private Text Rewriting (2022.coling-1)
Copied to clipboard
| Challenge: | Existing systems for differentially private text rewriting lack the means to validate privacy-preserving claims. |
| Approach: | They propose an open-source framework for differentially private text rewriting which is modular, extensible and highly customizable. |
| Outcome: | The proposed framework provides a way to lead and validate private text rewriting research. |
Privacy-Preserving Natural Language Processing (2023.eacl-tutorials)
Copied to clipboard
| Challenge: | This tutorial will help the NLP community to get familiar with current research in privacy-preserving methods. |
| Approach: | This tutorial will help the NLP community to get familiar with current research in privacy-preserving methods. |
| Outcome: | The tutorial will cover membership inference, differential privacy, homomorphic encryption, or federated learning, all with typical use-cases and potential pitfalls. |
Differentially Private Natural Language Models: Recent Advances and Future Directions (2024.findings-eacl)
Copied to clipboard
| Challenge: | Recent advances in deep learning have led to great success in various natural language processing tasks. |
| Approach: | They propose a systematic review of recent advances in DP deep learning models . they discuss some differences and additional challenges of DP-NLP . |
| Outcome: | The proposed method can prevent reconstruction attacks and protect against potential side knowledge while maintaining the privacy of sensitive data. |
When differential privacy meets NLP: The devil is in the detail (2021.emnlp-main)
Copied to clipboard
| Challenge: | Differential privacy provides a formal approach to privacy of individuals. |
| Approach: | They propose to use ADePT to provide differentially private auto-encoders for text rewriting to provide tight privacy guarantees for users' original utterances. |
| Outcome: | The proposed algorithm is not differentially private, thus rendering the experimental results unsubstantiated. |
Answering legal questions from laymen in German civil law system (2024.eacl-long)
Copied to clipboard
| Challenge: | Existing studies have focused on questions asked by experts, such as lawyers or legal scholars. |
| Approach: | They use a dataset to analyze laymen's legal questions paired with answers from lawyers and grounded to concrete law book paragraphs to find out what limitations exist. |
| Outcome: | The proposed system could help laymen in real situations without understanding law . the proposed system is based on 21k laymen’s legal questions paired with answers from lawyers and grounded to concrete law book paragraphs. |
To Share or Not to Share: What Risks Would Laypeople Accept to Give Sensitive Data to Differentially-Private NLP Systems? (2024.lrec-main)
Copied to clipboard
| Challenge: | Existing studies on differential privacy in NLP mainly focus on technical aspects, neglecting the human perception of privacy risks. |
| Approach: | They propose to use a differentially private algorithm to determine the privacy budget to determine which values are acceptable in which situations. |
| Outcome: | The proposed study aims to determine what thresholds would lead lay people to share sensitive textual data. |
The Impact of Inference Acceleration on Bias of LLMs (2025.naacl-long)
Copied to clipboard
| Challenge: | Recent work suggests strategies to increase inference efficiency with LLMs . however, these strategies may inadvertently lead to some side-effects. |
| Approach: | They propose to optimize inference acceleration strategies such as quantization, pruning, and caching to reduce inference cost and latency while maintaining predictive performance. |
| Outcome: | The proposed strategies reduce cost and latency while maintaining predictive performance while preserving the model size. |
Private Synthetic Text Generation with Diffusion Models (2025.naacl-long)
Copied to clipboard
| Challenge: | Recent research shows diffusion models are capable of generating synthetics texts . but are they also good in generating private data if the training was under differential privacy? |
| Approach: | They examine whether diffusion models are capable of generating synthetics texts . they critically assess previous work on private text generation with diffusion models . |
| Outcome: | The proposed model outperforms auto-regressive models in generating private images despite unmet privacy assumptions . the proposed model is open-source and can be used for other purposes . |
One size does not fit all: Investigating strategies for differentially-private learning across NLP tasks (2022.emnlp-main)
Copied to clipboard
| Challenge: | Existing research on the efficiency of differentially-private stochastic gradient descent (DP-SGD) in NLP is inconclusive or even counter-intuitive. |
| Approach: | They propose to use differentially-private stochastic gradient descent (DP-SGD) to preserve privacy in NLP by using modern neural models based on BERT and XtremeDistil architectures to conduct extensive experiments. |
| Outcome: | The proposed models and training strategies provide the best trade-off between privacy and performance on different NLP tasks. |