Papers by Mikaela Keller
To Word Senses and Beyond: Inducing Concepts with Contextualized Language Models (2024.emnlp-main)
Copied to clipboard
| Challenge: | Word Sense Disambiguiation and Word sense Induction are considered independent problems, but they are often neglected in practice. |
| Approach: | They propose an unsupervised task of learning a soft clustering amongwords that defines a set of concepts directly from data. |
| Outcome: | The proposed approach leverages both a local and global cross-lexicon view to induce concepts and also senses in the context of the proposed task. |
Fair NLP Models with Differentially Private Text Encoders (2022.findings-emnlp)
Copied to clipboard
| Challenge: | Encoded text representations often capture sensitive attributes about individuals, raising privacy concerns and making models unfair to certain groups. |
| Approach: | They propose an approach that combines privacy and adversarial training to learn private representations which induces fairer models. |
| Outcome: | The proposed approach improves on four NLP datasets and shows that privacy and fairness can positively reinforce each other. |
A Tale of Two Laws of Semantic Change: Predicting Synonym Changes with Distributional Semantic Models (2023.starsem-1)
Copied to clipboard
| Challenge: | Lexical Semantic Change is the study of how the meaning of words evolves through time. |
| Approach: | They propose to use distributional models to detect whether LD or LPC operate for given word pairs. |
| Outcome: | The proposed frameworks achieve a balanced accuracy above 0.6 on the dataset. |
Fair Without Leveling Down: A New Intersectional Fairness Definition (2023.emnlp-main)
Copied to clipboard
| Challenge: | Existing approaches to capture intersectional group fairness lack significant unfairness at intersection levels. |
| Approach: | They propose a new definition of intersectional fairness that combines absolute and relative performance across sensitive groups. |
| Outcome: | The proposed definition does not improve on a simple baseline. |