Topic Modeling of Everyday Sexism Project Entries

Abstract

The Everyday Sexism Project documents everyday examples of sexism reported by volunteer contributors from all around the world. It collected 100,000 entries in 13+ languages within the first 3 years of its existence. The content of reports in various languages submitted to Everyday Sexism is a valuable source of crowdsourced information with great potential for feminist and gender studies. In this paper, we take a computational approach to analyze the content of reports. We use topic-modelling techniques to extract emerging topics and concepts from the reports, and to map the semantic relations between those topics. The resulting picture closely resembles and adds to that arrived at through qualitative analysis, showing that this form of topic modeling could be useful for sifting through datasets that had not previously been subject to any analysis. More precisely, we come up with a map of topics for two different resolutions of our topic model and discuss the connection between the identified topics. In the low resolution picture, for instance, we found Public space/Street, Online, Work related/Office, Transport, School, Media harassment, and Domestic abuse. Among these, the strongest connection is between Public space/Street harassment and Domestic abuse and sexism in personal relationships.The strength of the relationships between topics illustrates the fluid and ubiquitous nature of sexism, with no single experience being unrelated to another.

Semi-supervised Multi-task Learning for Multi-label Fine-grained Sexism Classification

“…We end this section with a brief review of multi-task learning. Melville et al (2018) apply topic modeling to data obtained from The Everyday Sexism Project and maps the semantic relations between topics. ElSherief et al (2017) study user engagement with posts related to gender based violence and their language nuances.…”

A culpa é dela! É isso o que dizem nos comentários das notı́cias sobre a tentativa de feminicı́dio de Elaine Caparroz

“…Foram verificadas as seguintes opções para construção de tópicos e palavras (T-P), respectivamente: 3-5; 3-10; 5-3; 5-5; 5-10; 10-5; 10-10. Esta parametrização baseou-se em trabalhos que realizaram modelagem de tópicos para dados de mídias sociais como ] e [Melville et al 2019] e também na experiência dos pesquisadores. Ao alternar entre os parâmetros disponíveis, ou a dispersão entre esses ficava muito grande, o que ocasionava uma grande repetição de tópicos ou palavras; ou então a quantidade e composições eram insuficientes para permitir a rotulação de todos.…”

“…O grupo "Internet/Social", por exemplo, pode ser associado a comentários que criticam a atitude de Elaine Caparroz ao utilizar as redes sociais para um encontro, exemplo citado no Item 6 da Tabela 4. A abordagem utilizada foi semelhante ao de [Melville et al 2019].…”

“…Para a modelagem de tópicos foram utilizados o Latent Semantic Analysis (LSA) e o Latent Dirichlet Allocation (LDA), tal como em] e[Melville et al 2019]. Os dois métodos foram escolhidos por serem bem aceitos pela comunidade e estarem presentes em diversos trabalhos que visam analisar dados advindos de redes sociais.…”

Co-designing Cards on Social Issues for Creating Educational Games

“…The participants were also asked to read, individually, 8 written lived experiences related to the same category of everyday sexism and select the 3 most representative ones in groups. The lived experiences were extracted from a website called everydaysexism.com where people shared their experiences on everyday sexism (Melville et al, 2017). Reflective questions aimed at triggering reflection by future card users who would not necessarily be aware on the issues.…”