top of page

Scalable Gender Balance Analysis of Street Names in Cities Using Multi-Sensor Data and AI

Writer's picture: LLcloud NewsLLcloud News

Updated: Jan 14



Past research has clearly shown the lack of gender balance in street names in major cities in Europe, the USA and India. The Mapping Diversity project by Italian and European journalists shows the results of the analysis of 145,933 streets across 30 major European cities, located in 17 different countries. More than 90% of the streets named after individuals are dedicated to white men. This is the first large-scale journalistic project covering the street names of dozens of cities in many European countries. Various other smaller journalistic projects have analysed street names from different national or local perspectives in Spain, Germany, Hungary and Portugal.


Another notable previous study was done by a group of researchers and software developers at Mapbox lead by Aruna Sankaranarayanan. The group mapped seven cities: London, Paris, San Francisco, Mumbai, New Delhi, Chennai, and Bangalore. They found that, on average, only 27.5 percent of the studied streets had female names.


Some gender analysis results of street names from the Mapping Diversity project about the city of Lisbon are shown below.






While the Mapping Diversity approach for street name gender classification using Wikipedia and human curation produces some great results, it is clearly not scalable to cover many continents and cities, nor is it suitable for continuous monitoring of street name changes.


An alternative approach using AI platforms and APIs with AI trained models for name gender analysis, e.g. https://genderize.io/ and https://gender-api.com/ was designed and used with the aid of the LLcloud platform by an international group of researchers from the 'Dynamics of Placemaking' COST Action instead to deal with these shortcomings. The group requested and received gender classified street name data in .geojson format from the Mapping Diversity project for the following cities Athens, Lisbon, Vienna, Copenhagen, Madrid, Kiyv, selected to cover different languages and alphabets and having a good geographical spread in Europe. These city datasets have been used in LLcloud for comparison purposes between the Wikipedia and human curation approach applied by the Mapping Diversity project team and the AI model gender classification of street names



The https://genderize.io/ statistical/AI name gender classification platform and API was used for this study. To analyse objectively what is the potential of using such approaches and tools for automated street name gender classification, only street names of people from the Mapping Diversity data were pre-processed & analysed with the https://genderize.io/ and compared..


Using https://genderize.io/ AI model with pre-processing for name gender analysis.  Streets are marked in green where the two approaches gave the same gender classification results and in red where the classification results differed.
Using https://genderize.io/ AI model with pre-processing for name gender analysis. Streets are marked in green where the two approaches gave the same gender classification results and in red where the classification results differed.

Overall, results of between 70% to 90% agreement between the Mapping Diversity classification approach and the new AI name gender classification approach have been obtained for the cities of Lisbon, Vienna and Copenhagen.


To further validate in situ the AI gender classification results, the photo Spot mobile app integrated with the LLcloud platform was used to capture street name signs in Lisbon and overlay them with the street names from Open Street Maps in LLcloud and the AI classification results.



The long-term goal and ambition is to create a scalable service in LLcloud with global coverage, which will allow placemaking researchers, city authorities, governments and international organisations to study and monitor continuously for any city around the world the gender balance and diversity of street names.




9 views0 comments

Comments


Personal Data Protection (EN/BG)

Copyright © 2022- by LLcloud Ltd.

bottom of page