Krithika Ramesh

Hey! I’m Krithika, a Ph.D. student in Computer Science at the Johns Hopkins Centre for Language and Speech Processing where I’m advised by Dr. Anjalie Field. My research centers primarily on privacy and memorization in NLP, and other areas that are essential for the development of trustworthy NLP systems, such as interpretability and fairness.

I’ve previously worked as a research intern at Microsoft Research India under the guidance of Dr. Sunayana Sitaram. Prior to this, I interned with the EzPC team at Microsoft Research India, and at the Indian School of Business, where I was mentored by Dr. Sumeet Kumar and Dr. Ashique KhudaBukhsh.

If you’re still reading this, you probably (maybe, hopefully) also want to know a little about me as a person. I run, I read (to some extent), I love video games (though I don’t play them as much anymore), specifically RPGs like Divinity Original Sin: II, Mass Effect and Dragon Age. I’ve even actually recently begun to learn to play the keyboard so that I can play the music from these games! Some of the podcasts that I think are fantastic are Unexplainable and Flash Forward (does this page sound more like an advert at this point?).

I’m always open to discussion, so if you have any queries related to anything, or if there’s any way I could help you, feel free to drop me a mail or contact me via Twitter!

news

Sep 20, 2024	Our paper Evaluating Differentially Private Synthetic Data Generation in High-Stakes Domains was accepted to the Findings of EMNLP 2024!
May 2, 2023	Our paper A Comparative Study of Model Compression Techniques on Fairness in Language Models was accepted to ACL 2023!
Apr 15, 2023	I’ll be starting my Ph.D. at Johns Hopkins CLSP this fall!
Jan 21, 2023	Our paper Fairness in Language Models Beyond English: Gaps and Challenges was accepted to the Findings of EACL 2023!

selected publications

Evaluating Differentially Private Synthetic Data Generation in High-Stakes Domains

Krithika Ramesh, Nupoor Gandhi, Pulkit Madaan, Lisa Bauer, Charith Peris, and Anjalie Field

In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing (Findings), Nov 2024

Abs

The difficulty of anonymizing text data hinders the development and deployment of NLP in high-stakes domains that involve private data, such as healthcare and social services. Poorly anonymized sensitive data cannot be easily shared with annotators or external researchers, nor can it be used to train public models. In this work, we explore the feasibility of using synthetic data generated from differentially private language models in place of real data to facilitate the development of NLP in these domains without compromising privacy. In contrast to prior work, we generate synthetic data for real high-stakes domains, and we propose and conduct use-inspired evaluations to assess data quality. Our results show that prior simplistic evaluations have failed to highlight utility, privacy, and fairness issues in the synthetic data. Overall, our work underscores the need for further improvements to synthetic data generation for it to be a viable way to enable privacy-preserving data sharing.
Fairness in Language Models Beyond English: Gaps and Challenges

Krithika Ramesh, Sunayana Sitaram, and Monojit Choudhury

Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics (Findings), May 2023

Abs

With language models becoming increasingly ubiquitous, it has become essential to address their inequitable treatment of diverse demographic groups and factors. Most research on evaluating and mitigating fairness harms has been concentrated on English, while multilingual models and non-English languages have received comparatively little attention. This paper presents a survey of fairness in multilingual and non-English contexts, highlighting the shortcomings of current research and the difficulties faced by methods designed for English. We contend that the multitude of diverse cultures and languages across the world makes it infeasible to achieve comprehensive coverage in terms of constructing fairness datasets. Thus, the measurement and mitigation of biases must evolve beyond the current dataset-driven practices that are narrowly focused on specific dimensions and types of biases and, therefore, impossible to scale across languages and cultures.