| Challenge: | Word embeddings such as word2vec and GloVe exhibit human-like implicit biases based on gender, race, and other social constructs. |
| Approach: | They propose a simple generaliza test to measure bias in word embeddings by comparing two sets of target-concept words to two sets . |
| Outcome: | The proposed test shows that word2vec and word2Ve exhibit human-like implicit biases based on gender, race, and other social constructs. |
Similar Papers
Sense Embeddings are also Biased – Evaluating Social Biases in Static and Contextualised Sense Embeddings (2022.acl-long)
Copied to clipboard
| Challenge: | Existing studies have evaluated social biases in word embeddings, but they are understudied. |
| Approach: | They propose to evaluate the social biases in sense embeddings using a benchmark dataset for word embedders. |
| Outcome: | The proposed measures show that even when no biases are found at word-level, there are still worrying levels of social biase at sense-level which are often ignored by the word- level bias evaluation measures. |
When do Word Embeddings Accurately Reflect Surveys on our Beliefs About People? (2020.acl-main)
Copied to clipboard
| Challenge: | a study of word embeddings shows that social biases are more accurate than survey data for some dimensions of meaning. |
| Approach: | a new study investigates the extent to which word embeddings accurately reflect biases . they find that biased word embeds mirror survey data across 17 dimensions of social meaning . |
| Outcome: | a new study shows that word embeddings accurately reflect biases on average across dimensions of social meaning . biased embedders are more reflective of survey data for some dimensions of meaning than others, the study finds . |
Robustness and Reliability of Gender Bias Assessment in Word Embeddings: The Role of Base Pairs (2020.aacl-main)
Copied to clipboard
| Challenge: | Existing methods to quantify gender bias in word embeddings are not robust and cannot identify common types of bias. |
| Approach: | They propose to quantify gender bias by using cosine similarity to a pair of gender words and using analogies. |
| Outcome: | The proposed methods are not robust and cannot identify common types of bias, while analogies are unsuitable indicators. |
Mind Your Bias: A Critical Review of Bias Detection Methods for Contextual Language Models (2022.findings-emnlp)
Copied to clipboard
| Challenge: | Existing methods for detection of biases in contextual language models are inconsistent and inconclusive. |
| Approach: | They propose to use word embedding association test to detect biases in contextual language models to compare them with other methods. |
| Outcome: | The proposed methods are inconsistent and inconclusive for language models with word embeddings. |
Investigating the Frequency Distortion of Word Embeddings and Its Impact on Bias Metrics (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Recent research has shown that static word embeddings can encode words’ frequencies, but little has been studied about this behavior. |
| Approach: | They propose to use static word embeddings to encode words' frequencies and to assess the impact of this relationship on embeddable bias metrics. |
| Outcome: | The proposed model shows that word embeddings can produce higher similarity between high-frequency words than other embeddables. |
Measuring Social Biases in Grounded Vision and Language Embeddings (2021.naacl-main)
Copied to clipboard
| Challenge: | Existing methods to measure social biases in word embeddings are limited to visually grounded word embeds . a new study generalizes word embedment associations to visually ground word embeddas . |
| Approach: | They generalize word embeddings' biases to visually grounded word embeds . they propose two generalizations that answer questions about how biase, language, and vision interact . |
| Outcome: | The proposed measures are applied to a new dataset that includes 10,228 images from COCO, Conceptual Captions, and Google Images. |
SOS: Systematic Offensive Stereotyping Bias in Word Embeddings (2022.coling-1)
Copied to clipboard
| Challenge: | Systematic Offensive Stereotyping (SOS) in word embeddings could lead to associating marginalised groups with hate speech and profanity. |
| Approach: | They propose a quantitative measure of the systematic offensive stereotyping (SOS) in word embeddings and validate it in most commonly used word embeds. |
| Outcome: | The proposed measure correlates with published statistics on online extremism, but does not explain hate speech detection models. |
Unpacking Bias: An Empirical Study of Bias Measurement Metrics, Mitigation Algorithms, and Their Interactions (2024.lrec-main)
Copied to clipboard
| Challenge: | Word embeddings (WE) models reflect gender, racial, and religious stereotypes from the corpus on which they are trained. |
| Approach: | They propose a method that carefully controls for word sets and vector normalization to address these factors. |
| Outcome: | The proposed method detects consistency between different mitigation methods and the evaluation words used by the mitigation methods. |
On the Interpretability and Significance of Bias Metrics in Texts: a PMI-based Approach (2023.acl-short)
Copied to clipboard
| Challenge: | Word embeddings have been used to quantify biases in texts for years, but their statistical properties and advantages have not been studied. |
| Approach: | They propose to use PMI-based metric to quantify bias in corpora by conditional probabilities and odds ratio to approximate it. |
| Outcome: | The proposed measure can be approximated by an odds ratio, which makes statistical inferences cost-effective and meaningful. |
Representation biases in sentence transformers (2023.eacl-main)
Copied to clipboard
| Challenge: | argued that transformer-based models are not well suited for sentence-level downstream tasks. |
| Approach: | They propose to use sentence transformers to produce full-sentence representations . they propose to combine transformers with a training regime that embeds tokens into the model . |
| Outcome: | The proposed model performs better on downstream tasks than the vanilla model and its variants. |