Social media messages are taking a central role in spreading hate speech against women. Insults, threats, verbal attacks published on social media pages, and in online comments are showcases of aggressions moved on the net without filters and controls. According to VOX (Italian observatory on rights) 2020, women remain the real target for discriminatory messages: 63.1% of negative tweets are in fact directed against them.
A form of violence that seems unstoppable, as evidenced by the facts that many news continue to report incessantly. The sentiment analysis enables analysts to associate a polarity (positive or negative) to the Tweets published on the web, on the basis of the meaning of each Tweet content. From a methodological point of view, it seems very difficult to identify the hate speech, and it is mainly due to two factors: (i) the ambiguity of some language terms contained in the Tweets with a reference lexicon, (ii) the need to contextualize and understand the neutral, the false negative or positive opinions conveyed by the Tweets analyzed. In fact, terms that may indicate the presence of the hate speech are often used even without that purpose (we trivially think of ironic Tweets).
The use of innovative techniques has allowed us to correctly label the hate speech even in the absence of a precise aggressive lexicon. The expressions, e.g., “good woman”, and “available woman” might automatically produce a positive label, but in current use, it is synonymous with “prostitute”. The same when we use some masculine expressions that may take on a positive meaning. In order to address to this severe limit of the sentiment techniques, the production of lists of expressions, not just words, describing in a more precise way different nuances of sentiments becomes essential for the disambiguation process. In fact, automatic classification is not sufficient, if it is not proceeded by a phase of construction of an appropriate lexical dictionary.
This methodology allows to weight words and takes into consideration their mutual association and combination in syntagms that restitute a clearer meaning of their use in a specific context. Other factors are also considered, such as the average number of Tweets coming from a specific area or the spread of users in that particular area of the country.
Violence has many faces, not only that of men who openly attack women, but also the one of those who creep words in a subtle way. In both cases, messages of violence are persistently whispered, published on the Net, and applauded by the vast audience of social media. A study commissioned by the European Parliament for the Commission on Women’s Rights and Gender Equality has shown that one in ten women in Europe has experienced some form of cyber violence since the age of 15.
However, you cannot fight what you cannot even define: The United Nations and the European institutions recognize the existence of cyber violence (that perpetrated through social media and the Internet), and hate speech (the language of hatred) against women, even if a commonly accepted definition is still not available.
Building a gender sentiment analysis dictionary could be the first step in setting the boundary beyond which words become violence and abuse, because words can hurt more than a slap.
Read the full article here.