Learning Gender-Neutral Word Embeddings
arXiv:1809.01496
Abstract
Word embedding models have become a fundamental component in a wide range of Natural Language Processing (NLP) applications. However, embeddings trained on human-generated corpora have been demonstrated to inherit strong gender stereotypes that reflect social constructs. To address this concern, in this paper, we propose a novel training procedure for learning gender-neutral word embeddings. Our approach aims to preserve gender information in certain dimensions of word vectors while compelling other dimensions to be free of gender influence. Based on the proposed method, we generate a Gender-Neutral variant of GloVe (GN-GloVe). Quantitative and qualitative experiments demonstrate that GN-GloVe successfully isolates gender information without sacrificing the functionality of the embedding model.
EMNLP 2018
References in corpus (7)
- Efficient Estimation of Word Representations in Vector Space
- Semantics derived automatically from language corpora contain human-like biases
- Word Embeddings Quantify 100 Years of Gender and Ethnic Stereotypes
- A large annotated corpus for learning natural language inference
- Men Also Like Shopping: Reducing Gender Bias Amplification using Corpus-level Constraints
- Beyond Bilingual: Multi-sense Word Embeddings using Multilingual Context
- Bilingual Learning of Multi-sense Embeddings with Discrete Autoencoders
Cited by in corpus (15)
- Measuring and Reducing Gendered Correlations in Pre-trained Models
- Hurtful Words: Quantifying Biases in Clinical Contextual Word Embeddings
- Evaluating the Underlying Gender Bias in Contextualized Word Embeddings
- Interpreting Deep Learning Models in Natural Language Processing: A Review
- Reducing Sentiment Bias in Language Models via Counterfactual Evaluation
- Measuring Bias in Contextualized Word Representations
- What's in the Box? A Preliminary Analysis of Undesirable Content in the Common Crawl Corpus
- Dynaboard: An Evaluation-As-A-Service Platform for Holistic Next-Generation Benchmarking
- Exposing and Correcting the Gender Bias in Image Captioning Datasets and Models
- Understanding Undesirable Word Embedding Associations
- A Causal Inference Method for Reducing Gender Bias in Word Embedding Relations
- Compass-aligned Distributional Embeddings for Studying Semantic Differences across Corpora
- Removing Spurious Features can Hurt Accuracy and Affect Groups Disproportionately
- Automatically Inferring Gender Associations from Language
- Towards classification parity across cohorts