AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
Efficient Metadata Generation and Customized Solutions
The speaker highlights how their company used coactive and dynamic tagging to codify the community guidelines and processes, significantly reducing manual labeling efforts by 85%. They were able to deliver projected results in just two months instead of six, greatly reducing the amount of data that needed manual annotation. The powerful aspect of their solution is the ability to customize for different fandoms, which existing AI-powered solutions couldn't provide. By using weak supervision and semi-supervision techniques, the company was able to address challenges in programmatic labeling for discrete text problems and employ active learning to quickly provide accurate predictions.