Over 80% of social media posts that mention Roma have offensive content

What is the public image of Roma in social media? Do they create conditions for discrimination and use hate speech? What statements, stereotypes and prejudices are most often used? A study from the end of 2023 provides up-to-date information on these questions. For the presentation of the survey and its results, on 11.01.2024 at 12:00. Amalipe Center will organize a press conference in Kepler Hall in Launchee (Sofia). It will be attended by Dimitar Dimitrov – data analyst, Boyan Zahariev (author of the report summarizing the results) and Deyan Kolev ( Chairman of Center Amalipe).

The summarized results show that at least 80% of the social media materials in Bulgarian language mentioning Roma in some form have offensive and prejudiced content. At least half of the materials show direct hatred with the use of offensive and derogatory expressions.

The material for the analysis was collected from social media between 1st September and 30th November 2023. Of the 1,252 texts mentioning Roma in some form, more than 80% had offensive content about Roma. The majority of the Bulgarian-language material on Roma is from Facebook, the material on this topic found on other social media is sporadic.
The period of data collection partly coincides with the election campaign for local elections in 2023. Therefore, it is not surprising that 120 (about 10%) of the materials dealt with the topic of voting in local elections, while 76 mentioned other topics related to the problems of the municipalities that became the subject of the election campaign. In September 2023, in the run-up to the start of the election campaign, local election issues are already prominent in the media and public debate. The influence of the atmosphere created by the local elections is also felt in other publications referring to the Roma, which do not deal directly with the elections themselves or with the problems of the municipalities. Quite a few texts are devoted to the mayoral race in the capital, where, in addition to the context of electoral fraud and social and housing problems in Roma neighbourhoods, Roma are present both in a positive and in a very negative way in the comments about the ethnic origin of one of the main candidates for mayor. The range of texts devoted to the municipal elections as a whole contains numerous examples of abusive speech towards Roma as a group and towards individuals. There are also numerous examples of the use of isolated names of Roma as direct insults to people who have never declared themselves to be of Roma origin, including those whom nobody considers to be Roma. In this use, the words in question are used as general term abuse, similar to foul language, in relation to both political opponents and other hated or disliked individuals, to express feelings of superiority, contempt and other extreme forms of rejection.

In 54% of the texts, the self-designation of the ethnic group Roma is used in some form, in 37% the external designation Gypsies is used. Usually, the use of an ethnic group’s self-designation is a way of showing respect and is a manifestation of politeness and respect. The use of the term Roma does not at all imply that the respective posts/posts are positive. It is common to find pejorative phrases such as ‘the Roma brothers’. In fact, the term “Roma” is more often used to refer to sarcastic remarks, while gypsies is more often found in direct and unconcealed insults to Roma. Words with a directly offensive meaning such as m* are used much less frequently – in 2-3% of publications.
The search of publications was performed automatically using keywords pre-selected by experts. The publications were analyzed experimentally using artificial intelligence. The methodology and the algorithms used for the automatic analysis were developed by a consortium of organisations from the MRG-Hungary, the Institute of Human Rights-Slovakia, and the Amalipe Centre within the COUCH (Countering Online Antigypsyism and Cyber Hate) project. The project is funded by the Citizens, Equality, Rights and Values Programme (CERV).

This analysis presents data from a secondary semi-automated processing that relies on a critical review of algorithm performance. The aim is, on the one hand, to present more reliable information about offensive speech towards Roma in social media, and on the other hand, to identify weaknesses and typical gaps/errors in the algorithms used for automatic processing.

This publication reflects the views only of the authors, and the European Commission cannot be held responsible for any consequences arising from the re-use of this publication.

Contact person: Atanas Atanasov – 0882 540 209 – Email/E-mail: cor.amalipe.vt@abv.bg

The project is funded by the Citizens, Equality, Rights and Values Programme (CERV).