CodeLink: Using a Semantic Wiki for Code Documentation

Documentation maintainence is a difficult and costly process. Existing forms of software documentation exist independent from the code it describes. As a result, problems arise when source code evolves since there is no traceability between the documentation and code domains. CodeLink, a semantic wiki designed for code documentation attempts to address this problem. CodeLink provides a platform for establishing traceability links between developer documentation and code concepts. Traceability links between documentation and code can be explicitly created by the user through annotations, or inferred through the use of natural language analysis. CodeLink employs an ontology as its knowledge model and a semantic wiki as its user interface. Annotations on traceability relations in the ontology serve to connect natural language phrases with formal ontology concepts. In this paper, the motivation, design and implementation of CodeLink are described. Particular focus is placed on the natural language processing (NLP) component of CodeLink. The process in which the NLP component infers traceability links from the natural language text in the semantic wiki is described in depth. Several methods of querying and retrieving information from the knowledge model are discussed, followed by a summary of related and future work.