Version Detection in Spreadsheets Based on Headers Similarity

In almost every company, institution, business, or even for personal use people employ spreadsheets to perform a variety of tasks. Spreadsheet programming is the most commonly used end-user programming language. Spreadsheets users create new spreadsheets based on existing ones and reuse data layout and computational logic, so there are multiple versions of spreadsheets. End-users have no background of professional programming skills. That is why they are not well aware of the importance of version control and keeping the record of previous documents. So, the version information is usually missing, and it creates a lot of mess when some error is found in one spreadsheet that needs to be resolved in all of its other versions, and it becomes almost impossible when users deal with a very large number of spreadsheets. In this paper, we propose an automatic approach to cluster spreadsheets and find different versions of spreadsheets based on the similarity of headers present in spreadsheets. The header detection is not done in a conventional way that is present in literature, but we use font attributes in order to detect headers. We applied our approach on 40 spreadsheets randomly selected from spreadsheet corpora like Enron, FUSE, and EUSES. The evaluation results showed that the proposed approach could detect versions of spreadsheets with higher precision and recall rates of 82.24% and 78.48% respectively. The F1-Measure achieved is 80.08%.