VortexDraft: A blockchain smart contract auto-generation system based on named entity recognition

Abstract

As blockchain technology continues to evolve, its inherent decentralization, transparency, and immutability offer increasingly viable applications for smart contracts in the financial sector. Smart contracts autonomously execute and enforce transaction terms without the need for third-party intermediaries, thus enhancing both transaction efficiency and security. However, the creation of smart contracts generally requires advanced programming expertise, limiting their accessibility to non-technical users. To address this challenge, this paper introduces VortexDraft, an automated smart contract generation system leveraging Named Entity Recognition (NER). The system employs a user-configurable NER dataset to train a BERT (Bidirectional Encoder Representations from Transformers) model, facilitating the accurate extraction of key information from natural language inputs, which is then automatically translated into Solidity-compliant smart contract code. Experimental evaluations demonstrate that VortexDraft achieves over 99% accuracy in NER tasks, with the generated Solidity code aligning closely with the specifications outlined in natural language descriptions. This significantly enhances the efficiency and precision of smart contract generation, offering a robust solution for decentralized finance and other application areas.