Concepts to capture Relations between Project Documents using Ontologies

It is difficult to reveal erroneous assumptions in software projects due to mostly only implicit relations between project documents. Errors that result from these assumptions are expensive to fix in later project phases. Capturing relations between project documents helps to verify documents and detects potential erroneous assumptions early in a project, hence can avoid errors early that might be expensive to fix later. This thesis contributes an approach that uses the mechanisms of ontologies to verify project documents against requirements expressed in a controlled language. Natural language, a major component in mostly early phase documentation of software projects, is a source of ambiguity in requirements and project documents and can lead to misunderstandings. This thesis proposes a controlled natural language (CNL), based on a grammar definition, with a restricted vocabulary and sentence structure to formulate requirements with reduced ambiguity. In order to relate project documents, a generic format is required to compare the documents’ contents. This thesis describes an approach to transform the knowledge in project documents into an ontology. The ontology is built using semantically annotated productions in a CNL for requirements. Different types of semantic annotations are proposed and the semantic conflicts that can arise between them are discussed. Furthermore, a semantic meta-grammar is presented that allows the definition of grammars whose languages contain semantically annotated sentences. The concepts presented in this thesis are prototypically implemented in a tool which enables the user to write unambiguous requirements that are translated into ontology axioms. These axioms are used to capture relations between project documents that the user can import into the tool.