Automated Software Documentation Using Linguistically-Based Text Generation

Linguistically-based text generation techniques will be used to automate the generation of software documentation from CASE tools. Two types of software documentation are typically produced: (l) formal documentation describing the structure, interfaces, data types and other mathematical properties of software, and (2) semantic documentation describing the meaning and purpose of software components and actions. Our automation of formal documentation will make use of data about software structure obtainable from CASE tools. Semantic documentation will require either a specialized domain (either by application domain or by system type) or additional structured input from users. The modularity of our proposed text generator design and the linguistic sophistication of our underlying language model will allow the production of fluent, professional-quality documentation from a wide variety of information sources (difrerent CASE representations or user input), The proposed documentation generator framework will use text planning and grammatical knowledge to produce more fluent generated text than is possible with template-based approaches and have less need for augmentation by manually created free text. By decreasing the reliance on manual documentation we will guarantee better consistency between software documentation and program code via the specifications which guide the production of code.