Progress of Word Embedding in Code Generation

How to use natural language to generate program code directly has become the focus of researchers, and realizing machine's understanding of natural language is an important foundation to solve this problem. In the natural language processing system based on deep learning, word embedding model can encode words and sentences, which significantly improves the capability of a neural network to process text data. Through the analysis of word embedding model, this paper describes its vital role in code generation, elaborates the process of using word embedding and neural network to automatically generate program code, and finally looks forward to its development trend.