A Web 3.0-Based Trading Platform for Data Annotation Service With Optimal Pricing

Shu Yang, Yilei Zhang, Laizhong Cui, Bin Deng, Taoyuan Chen, Qingzhen Dong

Abstract: Annotating data is becoming increasingly important with the prevalence of machine learning. However, due to limited computing resources and high costs, most data owners prefer to outsource the task of data annotation to service providers with more computing resources and experience in annotation. While different parties are decentralized and selfish, it is important for them to reach consensus. We develop a Web 3.0-based trading platform for data annotation services that aims to ensure security and botain optimal prices during trading. Specifically, user information and trading records are stored in the blockchain to safeguard the platform's security. Moreover, we introduce a market mechanism that automates the transaction negotiation. In the market, a game theory-based pricing algorithm is proposed to maximize the whole utility of both parties. Finally, we conduct extensive experiments, and the results show that our system can significantly enhance the security level and user utility.