What is in a name? Mitigating Name Bias in Text Embeddings via Anonymization
Sahil Manchanda, Pannaga Shivaswamy
TL;DR
This work identifies a name bias in text embeddings where similarity is disproportionately driven by names rather than semantic content. It introduces an inference-time text anonymization technique that removes identity references while preserving themes, achieving training-free debiasing. Across multiple models and downstream semantic similarity tasks, anonymized embeddings demonstrate reduced name-driven bias and improved alignment with human judgments. The approach is practical for deployment but highlights the need for multilingual extensions and careful identity-preservation trade-offs in real-world use cases.
Abstract
Text-embedding models often exhibit biases arising from the data on which they are trained. In this paper, we examine a hitherto unexplored bias in text-embeddings: bias arising from the presence of $\textit{names}$ such as persons, locations, organizations etc. in the text. Our study shows how the presence of $\textit{name-bias}$ in text-embedding models can potentially lead to erroneous conclusions in assessment of thematic similarity.Text-embeddings can mistakenly indicate similarity between texts based on names in the text, even when their actual semantic content has no similarity or indicate dissimilarity simply because of the names in the text even when the texts match semantically. We first demonstrate the presence of name bias in different text-embedding models and then propose $\textit{text-anonymization}$ during inference which involves removing references to names, while preserving the core theme of the text. The efficacy of the anonymization approach is demonstrated on two downstream NLP tasks, achieving significant performance gains. Our simple and training-optimization-free approach offers a practical and easily implementable solution to mitigate name bias.