What is in a name? Mitigating Name Bias in Text Embeddings via Anonymization

2025-02-05

AI Summary

This paper identifies and mitigates the “name bias” problem in text embedding models: named entities—such as person names, locations, and organizational names—distort semantic similarity judgments, causing topically unrelated texts sharing such names to be erroneously deemed similar (or vice versa). To address this, we propose a lightweight, inference-time text anonymization method that leverages named entity recognition (NER) to identify and replace named entities, thereby preserving topical semantics without model retraining or parameter optimization. The approach is practical, interpretable, and deployment-friendly. Extensive experiments on clustering and retrieval tasks demonstrate significant performance gains, empirically validating anonymization’s robust suppression of name bias. Crucially, this work achieves the first targeted disentanglement of named entity interference within embedding spaces—enabling isolation of lexical identity effects from semantic content.

Technology Category

• Natural Language Processing: Ethics — Bias, Fairness, Transparency & Privacy
• Computer Vision: Bias, Fairness & Privacy
• Machine Learning: Privacy

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

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 $ extit{names}$ such as persons, locations, organizations etc. in the text. Our study shows how the presence of $ extit{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 $ extit{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.

Problem

• Mitigate name bias in text embeddings
• Propose text anonymization during inference
• Improve performance in NLP tasks

Innovation

• Mitigates name bias
• Uses text-anonymization technique
• Enhances NLP task performance