We believe that bringing awareness to the biases is the first step to making a change. Observing the gender graph project clearly reveals that the media commonly associates toxic words with women. We consume this media every day therefore subliminally inherit these biases. Much of our community believes that feminism isn’t relevant anymore as women and men have “equal rights”. Hopefully this data driven evidence will be proof of the disparities that exist in the way we perceive gender, and that we still have a long way to go.
would it be possible to correlate this with any form of authorship of the content being analysed?
Yes but will require extra processing on the training data and code adjustment of word embedding tool. In our case the word2vec output produce binary model where each word correspond with a numerical vector with no other attributes. Also, most often then not the trained dataset itself is not categorized by article or author.
The Gender Graph uses machine learning (word2vec) to reveal gender biases in specific media sources. Users can plot where words lie on a scale of (he) to (she), and observe the differences that exist in the way we perceive gender in the media.