On Debiasing Text Embeddings Through Context Injection
Not on the map. Why?
Abstract (arXiv)
Current advances in Natural Language Processing (NLP) have made it increasingly feasible to build applications leveraging textual data. Generally, the core of these applications rely on having a good semantic representation of text into vectors, via embedding models. However, it has been shown that these embeddings capture and perpetuate biases already present in text. While a few techniques have been proposed to debias embeddings, they do not take advantage of the recent advances in context understanding of modern embedding models. In this paper, we fill this gap by conducting a review of 19 embedding models by quantifying their biases and how well they respond to context injection as a mean of debiasing. We show that higher performing models are more prone to capturing biases, but are also better at incorporating context. Surprisingly, we find that while models can easily embed affirmative semantics, they fail at embedding neutral semantics. Finally, in a retrieval task, we show that biases in embeddings can lead to non-desirable outcomes. We use our new-found insights to design a simple algorithm for top k retrieval, where k is dynamically selected. We show that our algorithm is able to retrieve all relevant gendered and neutral chunks.
Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 27 of its 31 references that are in the map. First posted on arXiv 14 Oct 2024.
Builds on
27 of 31 references.
References · 27
- 2018BERT: Pre-training of Deep Bidirectional Transformers for Language Understandingcited by 40,204
- 2019RoBERTa: A Robustly Optimized BERT Pretraining Approachcited by 13,266
- 2021Evaluating Large Language Models Trained on Codecited by 8,053
- 2013Efficient Estimation of Word Representations in Vector Spacecited by 4,346
- 2016Enriching Word Vectors with Subword Informationcited by 1,578
- 2016Man is to Computer Programmer as Woman is to Homemaker? Debiasing Word Embeddingscited by 1,304
- 2021BEIR: A Heterogenous Benchmark for Zero-shot Evaluation of Information Retrieval Modelscited by 1,225
- 2016Semantics derived automatically from language corpora contain human-like biasescited by 894
- 2022BERTopic: Neural topic modeling with a class-based TF-IDF procedurecited by 779
- 2023Towards General Text Embeddings with Multi-stage Contrastive Learningcited by 705
- 2018Mitigating Unwanted Biases with Adversarial Learningcited by 663
- 2022MTEB: Massive Text Embedding Benchmarkcited by 623
15 more
- 2024NV-Embed: Improved Techniques for Training LLMs as Generalist Embedding Modelscited by 466
- 2021Sentence-T5: Scalable Sentence Encoders from Pre-trained Text-to-Text Modelscited by 462
- 2024LLM2Vec: Large Language Models Are Secretly Powerful Text Encoderscited by 396
- 2023Improving Text Embeddings with Large Language Modelscited by 295
- 2019Lipstick on a Pig: Debiasing Methods Cover up Systematic Gender Biases in Word Embeddings But do not Remove Themcited by 268
- 2023SugarCrepe: Fixing Hackable Benchmarks for Vision-Language Compositionalitycited by 238
- 2019Gender Bias in Contextualized Word Embeddingscited by 203
- 2019Black is to Criminal as Caucasian is to Police: Detecting and Removing Multiclass Bias in Word Embeddingscited by 146
- 2021Debiasing Pre-trained Contextualised Embeddingscited by 107
- 2023AnglE-optimized Text Embeddingscited by 98
- 2024SUGARCREPE++ Dataset: Vision-Language Model Sensitivity to Semantic and Lexical Alterationscited by 42
- 2018What are the biases in my word embedding?cited by 30
- 2019Debiasing Embeddings for Reduced Gender Bias in Text Classificationcited by 19
- 2024LangProp: A code optimization framework using Large Language Models applied to drivingcited by 17
- 2018An Analysis of Hierarchical Text Classification Using Word Embeddingscited by 16
Built on by
Not cited yet.