Papers by Padmini Srinivasan
A Girl Has A Name: Detecting Authorship Obfuscation (2020.acl-main)
Copied to clipboard
| Challenge: | Existing authorship attribution methods are not stealthy as they degrade text smoothness in detectable manner. |
| Approach: | They evaluate the stealthiness of authorship attribution methods under an adversarial threat model and show that they are not stealthy . |
| Outcome: | The proposed methods can be identified with an average F1 score of 0.87 . |
C3PA: An Open Dataset of Expert-Annotated and Regulation-Aware Privacy Policies to Enable Scalable Regulatory Compliance Audits (2024.emnlp-main)
Copied to clipboard
Maaz Musa, Steven Winston, Garrison Allen, Jacob Schiller, Kevin Moore, Sean Quick, Johnathan Melvin, Padmini Srinivasan, Mihailis Diamantis, Rishab Nithyanand
| Challenge: | Privacy policies fall short of achieving compliance goals due to their inaccessibility or incomprehensibility. |
| Approach: | They propose to use C3PA to create an open regulation-aware dataset of expert-annotated privacy policies to aid automated audits of compliance with CCPA-related disclosure mandates. |
| Outcome: | The proposed dataset is uniquely suited for aiding automated audits of compliance with CCPA-related disclosure mandates from 411 unique organizations. |
On the Robustness of Offensive Language Classifiers (2022.acl-long)
Copied to clipboard
| Challenge: | Existing studies on offensive language classifiers have focused on primitive attacks such as misspellings and extraneous spaces. |
| Approach: | They analyze the robustness of offensive language classifiers against crafty adversarial attacks that leverage greedy- and attention-based word selection and context-aware embeddings for word replacement. |
| Outcome: | The proposed classifiers are robust against more crafty attacks that leverage greedy- and attention-based word selection and context-aware embeddings for word replacement. |
Smells like Teen Spirit: An Exploration of Sensorial Style in Literary Genres (2022.coling-1)
Copied to clipboard
| Challenge: | Sensory perceptions shape how we use language and communicate. |
| Approach: | They examine the influence of sensorial language on writing style in a collection of lyrics, novels, and poetry. |
| Outcome: | The authors find that individual use of sensorial language is not a random phenomenon; choice is likely involved. |
LLM Bias Detection and Mitigation through the Lens of Desired Distributions (2025.emnlp-main)
Copied to clipboard
| Challenge: | Prior work on bias mitigation has focused on promoting social equality and demographic parity, but less attention has been given to aligning LLM’s outputs to desired distributions. |
| Approach: | They propose a weighted adaptive loss based fine-tuning method that aligns LLM’s gender–profession output distribution with the desired distribution while preserving language modeling capability. |
| Outcome: | The proposed method achieves near-complete mitigation under equality and 30–75% reduction under real-world settings. |
Don’t sweat the small stuff, classify the rest: Sample Shielding to protect text classifiers against adversarial attacks (2022.naacl-main)
Copied to clipboard
| Challenge: | Current text classifiers are subject to adversarial attacks from adversaries, typically executed using machine learning methods. |
| Approach: | They propose a novel and intuitive defense strategy called Sample Shielding that is attacker and classifier agnostic and does not require reconfiguration of the classifier or external resources. |
| Outcome: | The proposed defense is attacker and classifier agnostic and does not require reconfiguration of the classifier or external resources and is simple to implement. |
Debiasing by obfuscating with 007-classifiers promotes fairness in multi-community settings (2025.coling-main)
Copied to clipboard
| Challenge: | a number of studies have focused on the mitigation of biases in text classifiers. |
| Approach: | They propose an obfuscation-based data augmentation debiasing approach to reduce bias . they add to the training data *obfuses* versions of *all* false positive instances . |
| Outcome: | The proposed approach reduces bias for almost all of the tests without sacrificing false positive rates or F1 scores for minority or majority communities. |
Through the Looking Glass: Learning to Attribute Synthetic Text Generated by Language Models (2021.eacl-main)
Copied to clipboard
| Challenge: | Recent advances in natural language processing have enabled synthetic text generation that is often comparable to the organic text. |
| Approach: | They propose and test several ML-based methods to attribute authorship of synthetic text to language models (LMs) they propose to use a fine-tuned version of XLNet to achieve excellent accuracy . |
| Outcome: | The proposed method achieves excellent accuracy (91% to near perfect 98%) across a range of experiments where the synthetic text may be generated using pre-trained LMs, fine-tuned LM, or by varying text generation parameters. |
Suum Cuique: Studying Bias in Taboo Detection with a Community Perspective (2022.findings-acl)
Copied to clipboard
| Challenge: | Prior research has shown the need to consider community language norms when studying taboo text classification and annotations. |
| Approach: | They propose to use special classifiers tuned for each community's language to study bias in taboo classification and annotation where a community perspective is front and center. |
| Outcome: | The proposed method shows that biases are strongest against African Americans and South Asians . a community perspective is front and center in the proposed method . |
Adversarial Authorship Attribution for Deobfuscation (2022.acl-long)
Copied to clipboard
| Challenge: | Existing authorship attribution approaches do not consider adversarial threat model . authors show adversarially trained authorship attributors can degrade effectiveness of existing obfuscators from 20-30% to 5-10% . |
| Approach: | They propose to use rule-based and learning-based text obfuscation approaches to counter authorship attribution. |
| Outcome: | The proposed approaches do not consider the adversarial threat model . authors show that adversarially trained attributors can degrade effectiveness of existing obfuscators from 20-30% to 5-10% . |
Robust Bias Detection in MLMs and its Application to Human Trait Ratings (2025.findings-naacl)
Copied to clipboard
| Challenge: | Existing methods to assess demographic bias in MLMs ignore random variability of templates and target concepts, and neglect bias quantification. |
| Approach: | They propose a systematic statistical approach to assess bias in MLMs using mixed models to account for random effects, pseudo-perplexity weights for sentences derived from templates and quantify bias using statistical effect sizes. |
| Outcome: | The proposed method matches on bias scores in magnitude and direction with small to medium effect sizes. |