WIKIGENDER: A MACHINE LEARNING MODEL TO DETECT GENDER BIAS IN - - PowerPoint PPT Presentation

▶

Apr 11, 2023 421 likes •457 views

WIKIGENDER: A MACHINE LEARNING MODEL TO DETECT GENDER BIAS IN WIKIPEDIA Natalie Boln, Natlia Gulln, Sofia Kypraiou, Irene Petlacalco WIKIGENDER: METHODOLOGY Dataset Overviews from Wikipedia biographies: only 17% of them refer to women

SLIDE 1

WIKIGENDER:

A MACHINE LEARNING MODEL TO DETECT GENDER BIAS IN WIKIPEDIA

Natalie Bolón, Natàlia Gullón, Sofia Kypraiou, Irene Petlacalco

SLIDE 2

WIKIGENDER: METHODOLOGY

“Ada was an English mathematician and writer”

Stop word Adjective Noun

(0, 0, 1, 0, 0, …, 0, 1)

wordk ∈ overview wordn ∉ overview Binary target variable Train/Test partition

Logistic Regression

BALANCED DATASET PREDICTION & FEATURE EXTRACTION

for each occupation we use the same number of male and female entries

Encoding
Balancing dataset by occupation
Model

Overviews from Wikipedia biographies: only 17% of them refer to women

Dataset

https://wiki-gender.github.io

SLIDE 3

WIKIGENDER: RESULTS

Bias in Adjectives

women men beautiful

ffensive

profit certain cross hard creative defensive romantic diplomatic Top 5 most predictive adjectives Accuracy: 54.6±0.001%

Bias in Nouns + Adjectives

Accuracy: 62.9±0.002% Top 5 most predictive words women men person football marriage musician model

fficer

dancer war midfielder footballer Family Career Positive and strongly subjective Negative and weakly subjective

https://wiki-gender.github.io