Geography Markup Language (GML) becomes the de facto standard for geographic information representation on the
internet. GML schema provides a way to define the structure, content, and semantic of GML documents. It contains
useful structural information of GML documents and plays an important role in storing, querying and analyzing GML
data. However, GML schema is not mandatory, and it is common that a GML document contains no schema. In this
paper, we present Gstruct, a tool for GML schema extraction. Gstruct finds the features in the input GML documents,
identifies geometry datatypes as well as simple datatypes, then integrates all these features and eliminates improper
components to output the optimal schema. Experiments demonstrate that Gstruct is effective in extracting semantically
meaningful schemas from GML documents.