The Java Programmer's Phrase Book

Method names in Java are natural language phrases describing behaviour, encoded to make them easy for machines to parse. Programmers rely on the meaning encoded in method names to understand code. We know little about the language used in this encoding, its rules and structure, leaving the programmer without guidance in expressing her intent. Yet the meaning of the method names -- or phrases -- is readily available in the body of the methods they name. By correlating names and implementations, we can figure out the meaning of the original phrases, and uncover the rules of the phrase language as well. In this paper, we present an automatically generated proof-of-concept phrase book for Java, based on a large software corpus. The phrase book captures both the grammatical structure and the meaning of method phrases as commonly used by Java programmers.