Daniel starts with explaining what exactly he means with errors, defects and problems. According to IEEE the situation is like this:
A user makes an error, which results in a fault in the spreadsheet, leading to a failure.
For instance, if you enter a wrong number (error) there is a fault. But if no formulas use it, there is no failure, yet. All of these are anomalies, by the way.
The most prominent technique to detect anomalies is testing. For testing, you need to fill input cells with values and inspect output cells to see if they match your values.
So, for these test approaches, we need to detect what the user sees as input and output. This is what Daniel means with ‘cell meaning’. There are some problem with existing automatic approaches (that just consider cell relations)
- Checking formulas IF(A10=0,”ERROR”) are seen as output cells
- Aggregates mess with the situation: For example in a grading sheet, the results of the students are end-values you want to check, but if you average them, they do not show up as end-results anymore.
- Different user groups might have a different opinion on the meaning. Again for the grading sheet, the secretary might need the marks, but the professor is interested in the attendance rate only.