Information extraction for validation of software documentation

Abstract

Information extraction techniques can be used to improve the quality of software user manuals and online help systems. These documents are often formatted as repeated sections that have similar heading structure, with free-text inside each section. XML (extensible markup language) enables document designers to design rich tag sets where tags for section headings contain information about each section. This contextual information, coupled with the fact that the free-text portions of the documents use a limited sublanguage, mean that simple natural-language-based techniques can be used to extract facts from online documents. The SIFT document parser system has demonstrated the potential for this type of extraction in the area of software document validation.