Classifying and Recommending Knowledge in Reference Documentation to Improve Api Usability

Reference documentation is an important source of information on API usage. Programmers, however, can easily overlook reference information because of its tedious nature, and because the information they seek can be buried among irrelevant or boiler-plate text. We propose to detect and recommend fragments of API documentation relevant and important to a task. We categorize pieces of information in API documentation based on the type of knowledge they contain. From the pieces that contain knowledge worthy of recommendation, we extract the composition and the pattern of words, and use the patterns to automatically find new pieces that contain similar knowledge. In an evaluation study, with a training set of manually-classified reference documentation of about 1 000 API elements, we could issue recommendations with about, on average, 90% precision and 69% recall.