Hate speech is getting worse, which is a big problem on social media. This study looks at how machine learning (ML) can help reduce the bad effects of the internet by making it easier to moderate material. Hate speech is put into four categories: no hate, textit:intimidation, offense or discrimination, and promotes violence. This is done to give a more accurate range of intensity and get around the problems with looking at hate speech as a binary task, which is how sentiment analysis usually works. First, 31 people marked up a dataset in both English and German. The ML system has been put into a dashboard with explainable artificial intelligence (XAI) tools to make it easier to understand and be more open. A case study with 40 non-expert moderators is used to figure out how well the suggested XAI dashboard helps with content moderation. According to the results, it's important for content managers to know how bad hate speech is because it can affect how bad punishments might be. In a similar way, XAI seems like it could be a way to make ML more trustworthy, which would help judges make better decisions.
Here you can see the dashboard:
(A) Text Heatmapping (B) Confidence Barometer (C) Text Similarity (D) Evaluation
- User-friendly web interface for efficient moderation.
- Real-time storage of results using Firebase.
- Simultaneous collaboration for multiple users for the XAI dashboard.
- A dashboard that identifies hate speech and provides explanations for the machine learning model’s classification.
- Also Contains the User Case Study Groups A, B, C (More information about the project and the different groups can be found in the paper which will soon be public)
Install the required dependencies:
pip install -r requirements.txt
Obtain Firebase Credentials:
- Create a Firebase account.
- Create a Real-time Database and put the Credentials within the app.py
The code of the dashboard needs to be modified a lot, it is not yet directly usable. However, it contains all the important aspects to continue working on this dashboard. The interface of the dashboard and the main functions of the dashboard are public.
You are welcome to change this template to fit your project and its needs 😋
If you want to try the dashboard, don't hesitate to contact me to get access data to the dashboard. You can reach me at the e-mail address: Raisa.geleta@gmail.com
Check out the Hate Speech Detection XAI Dashboard to see what it looks like 🎉