Jan Krasni is a researcher and a lecturer focusing on multimodal discourse analysis, media studies (especially media ethics), and infrastructure studies. His work covers various fields of digital phenomena: starting with the construction of sentiments and meanings in the digital discourses, over-representation of minorities and minority languages in the media, and the problem of ethics in digital media infrastructure for processing personal information, up until the problem of fake news. Before joining the Technical University Berlin (Decoding Antisemitism project, where he leads the German team) he worked as a researcher at the Cultural Trends Lab with Lev Manovich at the University of Tyumen (Russia), and as an assistant professor at the Webster University.
Dr. Jan Krasni
All Sessions by Dr. Jan Krasni
-
Day 2
Sun, 29 May
-
11:30
Decoding Antisemitism Project: Labelling the Language or How to teach AI to recognise hate ideologies
11:30 - 12:00Hate ideologies and radicalisation on the web represent a new danger for societies around the world. These hateful standpoints are voiced often in the user comments on various platforms. However, it is physically impossible for humans to moderate the comment sections of the news websites and the social networks and remove the content spreading hate ideology. Usually, the moderators are using algorithms trained with the word and phrase lists characteristic of the right-wing or radical Islam milieu (in addition to various lists of slurs).
In Decoding Antisemitism Project, we experiment with the new approach in which we want to train the algorithms to recognise the hate ideologies on a conceptual basis. We work on antisemitism in user comments in three languages (French, English, German): which we are annotating (labelling) on a conceptual level. If our experiment will be successful, we will be able to showcase a completely new approach to natural language processing (NLP) within supervised machine learning. This talk will also demonstrate how insights from humanities can contribute to safety online if supported by machine learning.PresentationWhereGallery
-
DESCON started as a parody of DEF CON.
Our aim is to stay true.
For any questions, get in touch with us here.
Find us on Facebook, Twitter, Flickr or YouTube. -
-
Attendants will treat all within the DESCON community equally, irrespective of gender, nationality, racial or ethnic origin, religion or beliefs, disability, age, or sexual orientation and in accordance with ISOC Serbia Chapter PSEAH policy.
𝗗𝗮𝘁𝗮 𝗺𝗮𝗻𝗮𝗴𝗲𝗺𝗲𝗻𝘁 𝗮𝗻𝗱 𝗽𝗿𝗶𝘃𝗮𝗰𝘆
Your data will be securely processed and stored in compliance with the European General Data Protection Regulation 2016/679 and used only for the purposes of the registration. At any time you will have the possibility to consult, modify, update or delete your personal data by contacting the organizers at the following email address: info @ descon . me