Ontological Traceability using Natural Language Processing

Abstract

The software development process is continuously evolving to find the right balance between the real-life problem side with the requirements engineering and the architecture of the software as part of the solution.

While the software architecture has to deal with different types of artifacts using different notation, the requirement engineering has, on top, the complexity of the ambiguity of the real world as most of its artifact use natural language to capture the required functionality. To find a solution to link these two areas and the artifacts among them is one of the practical problems that look to narrow the gap between the expected solution and the actual solution.

The use of ontologies is an accepted theory as a solution to connect these two areas and their respective artifacts through the creation of trace links. Such trace links can track a requirement among the different artifacts. However, the effort to generate such trace links can be time consuming and not beneficial for the time of development.

This research project aims to propose a solution to automate the software traceability through the use of a conceptual representation of a software artifact. An artifact in natural language is represented as a sub-ontology and find its match in a Product Ontology using natural language techniques and tools.

To demonstrate such theory, a proof-of-concept is created to extract an ontology from a software artifact and fins trace match and trace links among other artifacts. The results when testing the concepts is remarkable and suitable, and the level of acceptance in a segment of the software industry is quite promising.