i-nth logo


Felienne Hermans, Martin Pinzger, & Arie van Deursen


The use of spreadsheets to capture information is widespread in industry. Spreadsheets can thus be a wealthy source of domain information. We propose to automatically extract this information and transform it into class diagrams. The resulting class diagram can be used by software engineers to understand, refine, or re-implement the spreadsheet's functionality.

To enable the transformation into class diagrams we create a library of common spreadsheet usage patterns. These patterns are localized in the spreadsheet using a two-dimensional parsing algorithm. The resulting parse tree is transformed and enriched with information from the library.

We evaluate our approach on the spreadsheets from the EUSES Spreadsheet Corpus by comparing a subset of the generated class diagrams with reference class diagrams created manually.


Class diagram extracted from the fruit example
Class diagram extracted from the fruit example

The information contained in the example spreadsheet can be represented by the accompanying class diagram.


2010, Delft University of Technology, Software Engineering Research Group, Report TUD-SERG-2010-013

Full article

Automatically extracting class diagrams from spreadsheets