Feature Location Using Extraction of Code Documentation

Feature location is a set of procedures of how to seek and find between software artifacts in pieces of source code. Code documentation is the breadcrumb of a puzzle which could help to get the developer's actual intention of specific code purpose. The token documents are extracted from the class name, method name, and Javadoc using natural language processing. The class name, method name, and Javadoc combine as documents. The VSM Lucene employed to index the document. The noun or verb of the use case scenario was transformed using a generic-specific relationship or replacement using the same meaning code name which was used as a query to identify the feature location. Each token from the use case scenario is categorized as class, method, or Javadoc to determine what is the best field to match. The query tokens are also preprocessed using NLP. The number of use case scenarios was 15, and the number of queries was about 150. The number of relevant files was 93, while the number of retrieved were 259. The number of relevant and retrieved was 65 files. As a result, the research could get the average precision rate was 24%, the average recall was 70%. The best precision was 50% and the best recall was 100%.