Duplicate content on a large site is rarely accidental copying. It is accumulated overlap, and the work is deciding what to merge rather than detecting it.
Overlap is a publishing-process problem
Four pages about the same topic exist because four people commissioned content over three years without a shared inventory. Fixing the pages without fixing the commissioning process guarantees the same state in another three years.
Merge into the URL with the history
When consolidating, keep the URL with the strongest link profile and longest history, not the one with the best content. Move the good content onto that URL and redirect the others. Teams routinely do this backwards and throw away the equity they were trying to consolidate.
Parameters and templates generate the bulk of it
Faceted navigation, session parameters and print views usually produce far more duplication than editorial overlap. That is a technical fix with canonical tags and parameter handling, not a content project, and it is normally the larger win.
Expect a temporary dip
Consolidation removes URLs, and indexed page counts fall. That is the intended outcome, and it looks alarming on a dashboard for several weeks. Tell whoever watches that dashboard before you start.