Time Period: 01.08.2023 - 31.01.2025
Funding Volume: 146.200,00€
Funded by: Hessian Ministry of the Interior, for Security and Homeland Security
Mittweida
Darmstadt
The BoTox project is dedicated to researching automated methods for bot and context recognition context recognition of hate comments on the internet. The aim is to protect social discourse and identify criminally relevant content, because in an increasingly networked digital space, the impact of hate speech and manipulative bots on our society poses a major challenge.
The automatic classification of hate comments has so far mostly been limited to recognizing hate comments, i.e. classifying a comment as a hate comment or a “normal” comment. The BoTox project goes beyond this binary classification.
Automated detection of criminally relevant hate speech is essential as it enables platform operators and law enforcement authorities to respond more effectively to criminally relevant content. In a space where freedom of expression and criminal law boundaries are a balancing act, it is crucial to develop tools that enable and support the identification and classification of hate comments according to relevant criminal offenses. German criminal law has 12 sections that are relevant to hate speech. In the BoTox project, hate speech is automatically analyzed to determine whether the content of the comments is relevant under criminal law and which paragraphs could apply. The aim is to provide platform operators and law enforcement authorities with assistance in managing the volumes and making a legal assessment.
Social bots that spread hate comments are particularly problematic, poisoning the discussion and damaging the democratic culture of debate. Such “hate bots”, which pretend to be people, influence the discourse either by automatically spreading content themselves and communicating with other users or by liking, sharing and commenting on hate speech. Both variants will be examined as part of the research project.
The detection of hate bots is extremely relevant for the Hessen3C Hate Comment Reporting Office for the prosecution of hate comments. In addition, if hate bots are detected, the networks can be asked to delete them or the public can be informed so that the manipulative goal can no longer be achieved.
Context detection in the context of hate comments is essential in order to understand and classify comments correctly, as comments are often misunderstood and cannot be evaluated correctly if they are viewed in isolation and the context is missing. The context detector aims to capture and analyze this context in order to enable a more precise evaluation of comments.
Up to now, hate speech detection and analysis has generally only looked at and evaluated individual comments taken out of context. However, comments can often only be interpreted correctly in a specific context. Especially when metaphors, irony or sarcasm play a role, the context can contribute to understanding.
BoTox aims to develop advanced methods for the automated detection of hate comments with criminal relevance. The focus is on the classification of criminally relevant hate comments, the development of tools for context analysis and the identification of manipulative hate bots. The aim is to provide platforms and authorities with effective means of taking action against hate speech and protecting the culture of discussion.
BoTox - Bot and context recognition in the context of hate comments
https://botox.h-da.de/