TikTok is a short-video platform where anybody can upload short videos. Uploated videos will not be checked on policy violations by moderators. Every user has the ability to report videos and comments that contain user claims. These reports identify content that needs to be reviewed by moderators. This process generates a large number of user reports, and the limited number of moderators makes it challenging to quickly address the reports.
Therefore TikTok needs a solution to reduce the backlog of user reports. I proposed to develop a predictive machine learning model that can determine whether a video contains a claim or offers an opinion. The dataset contains nearly 20000 reports that includes claim status of the report, video id, duration, view count, like count, share count, download count, comment count, transcription text, author verified status and author ban status.
To reach the goal, the development of the model is divided into following steps:
Note: The scenario and data in this project are entirely fictitious and do not represent TikTok or any real-world entities. This project is a modified version of the end-of-course projects developed as part of the Google Advanced Data Analytics Certificate Program.
You can try the final model and check a TikTok video by following this setup:
git clone https://github.com/mertyaylacigit/TikTok-Claims-Classification.git
cd TikTok-Claims-Classification
python3 -m venv venv
source venv/bin/activate
chmod +x ./install.sh
./install.sh
python3 script.py https://www.tiktok.com/@paralympics/video/7375186925335924001
Mert Yaylaci