Gender Inference Tools - GESIS Website
What is it about
The Gender Inference Gesis website provides a comprehensive overview and evaluation of existing methods to infer gender from (i) names, (ii) or (iii) mixed (names and images). Additionally, it provides a service as a form of API or downloadable package for non-commercial and academic use.
On this platform information is presented about:
- existing commercial and noncommercial services to infer gender from images, names or both
- the accuracy of gender inference tools, tested on different datasets. This comparison is also carried according to certain features (gender, age, ethnicity)
- instructions on how to use these services
This is a non-commercial service that is targeted at social scientists and other researchers from non-technical fields, who are
- interested in assessing gender attributes from images or names in their research and
- for whom reliable and unbiased demographic inferences are key.
While the gender inference methods we present focus on binary gender inference, we are aware that gender is better represented as a continuum than a binary valued concept. Many of the methods we present (?) output a continuous/confidence score which can be used to infer gender on a continuum. We hope with better training data available in the future with finer grained gender labels, it would be possible to extend/adapt these methods to infer gender on a continuum rather than binary values.