The University of Porto is developing a tool with artificial intelligence to help journalists, media and other online platforms detect and reduce hate speech on the Internet.
The researchers involved in the scientific project, called “Stop PropagHate“, collected during a week in January “thousands of news and wide tens of thousands of comments from the social network Twitter and various platforms associated with news collected in Portugal, UK, Brazil, and the USA, “said project coordinator Sérgio Nunes.
According to that expert, to develop the algorithm in question the team is extracting features of the news and studying how they influence the existence of comments with hate speech.
“We will study the characteristics that can be extracted automatically, that is, that do not require a human analysis, namely the media body in which the news is published, the category of news in the publication, the author of the news, theme s) of the news, people and places mentioned, words and sequences of words used in the text, number and diversity of adjectives used in the news and the time of publication, “explained Sérgio Nunes.
In the case of Portugal, the news gathering coincided with interviews with Mário Machado, a neo-Nazi several times convicted of racist crimes, who participated in a conversation about “We need a new Salazar?”, And in the case of Brazil, newsgathering for the investigation was linked to the election of the new president, Jair Bolsonaro.
In the UK, Brexit was chosen, while in the USA it was the shutdown and the issue associated with the wall on the Mexican border.
The project has a module focused on “automatic detection of hate speech in text” and another module focused on “detecting news that may result in an increase in hate speech” and “reactions with associated hate speech” a researcher at the University of Porto, in an interview in a preparatory work on ‘fake news’, the theme of a conference, to be held on February 21, in Lisbon, entitled “Fake News.
Using PropagHate’s Application Programming Interface (API), mass media can automatically identify hate speech within comments, but also predict the likelihood that a news story will generate those comments in the first place, “he adds.
Sérgio Nunes recalls that on Twitter and Facebook, the news that appear “are filtered” by platforms based on criteria that are not transparent.
“It is quite dangerous … to put so much power on opaque platforms … Why this order?”, Sérgio Nunes asks, noting that “most people are unaware that the news they see is leaked” and that are ordered according to “criteria that are neither transparent nor clear” and do not even appear in chronological order.
Fake news, commonly known as fake news, misinformation, or purposely falsified information for political or other purposes, gained prominence in the US presidents who elected Donald Trump in the referendum on Brexit in the United Kingdom and presidential in Brazil, won by the far-right candidate, Jair Bolsonaro.
The European Parliament wants to try to curb this phenomenon in the European elections in May and on 25 October 2018 adopted a resolution in which it advocates measures to strengthen the protection of personal data on social networks and combat the manipulation of elections following the abuse scandal personal data of millions of European citizens.
The “Stop PropagHate” project is being developed in the Department of Informatics of the Faculty of Engineering of the University of Porto (FEUP) in partnership with the Institute of Systems Engineering and Computers, Technology and Science (INESC-TEC), with the support of Digital News Initiative from Google.