1. David Doukhan - ddoukhan@ina.fr
GENDER EQUALITY MONITOR (GEM)
A large-scale Machine Learning approach for describing
representation and treatment differences existing between women
and men in media
2. Quantifying gender inequality in media is complex
How do people
speak?
What topics are covered?What are the wages?
Who decides?
Who is referred to?
3. Gender Equality Monitor project
● The project was initiated in July 2017 at INA
● Framework to describe gender representation differences in media
● Takes advantage of recent advances in AI and machine learning
● Fully automatic
● Multimodal: speaker, image and speech-to-text
● Speech-time analysis conducted on 1 million hours of audiovisual media
● Prospective results using face analysis and speech-to-text
● Softwares diffusion in open-source
● Estimates published in open-data
● This project has received funding from the European Union’s Horizon 2020
research and innovation programme under grant agreement No 780069
(MeMad project)
3
5. Titre de la présentation |12 janvier 2012 5
Men speak twice as much as women in French TV and radio!
Mean Women and Men speech-time from 2010 to 2018
6. Titre de la présentation |12 janvier 2012 6
Women speech-time percentage in French TV channels
7. Titre de la présentation |12 janvier 2012 7
Women speech-time percentage evolved on radio from 25.1% in
2001 to 34.4% in 2018
8. Multimodal estimation of women representation in French TV
Week of April 4th 2019
Women visual presence is higher than their vocal presence
Percentage of oral references to women is higher than women
speech time
8
10. Titre de la présentation |12 janvier 2012 10
To go further
Scientific publications
Doukhan, D., Carrive, J., Vallet, F., Larcher, A., & Meignier, S. (2018). An open-source speaker gender
detection framework for monitoring gender equality. In 2018 IEEE International Conference on
Acoustics, Speech and Signal Processing (ICASSP)
Doukhan, D., Poels, G., Rezgui, Z., & Carrive, J. (2018). Describing gender equality in french
audiovisual streams with a deep learning approach. VIEW Journal of European Television History and
Culture, 7(14).
Doukhan, D., Rezgui, Z., Poels, P., Carrive, J. (2019). Estimer automatiquement les différences de
représentation existant entre les femmes et les hommes dans les médias. journée DAHLIA :
”Informatique et Humanités numériques : quelles problématiques pour quels domaines ?”
Software
https://github.com/ina-foss/inaSpeechSegmenter
Data
https://www.data.gouv.fr/fr/organizations/institut-national-de-laudiovisuel