The current similarity measure for moved paragraphs is based on a percentage of changed and unchanged characters. We have implemented another version that is based on character runs instead, which might be a better measure for fragmentation.
Task
- Evaluate whether character percentages or character runs are the more effective approach, using the test pages in https://wmde-wikidiff2-unpatched.wmflabs.org/core/index.php/Main_Page
- Document the results either on the page or in the desktop document