Bayesian Unigram-Based Inference for Expanding Abbreviations in Source Code

Programmers often utilize abbreviations for naming identifiers when writing code. However, the difficulty to retrieve the original words of abbreviations during the maintenance phase makes the source code more problematic to comprehend. In this paper, we describe a Bayesian unigram-based inference to automatically expand abbreviations to their original words to enhance source code comprehension. Our approach automatically extracts a list of candidate words from the source code for a given abbreviation and employs the abbreviation's unigram statistical properties as evidence to find the best candidate word. We evaluated our approach on a set of 531 abbreviations randomly picked from eight open source projects and found that our approach correctly expands 83.62% of the set. Our approach provides an improvement of 48.3% in abbreviation expansion accuracy over the current state-of-the-art approaches.