Synthesis does a good job at automatically de-duplicating references. However, there are some extreme cases that we can not account for. The first is if an abstract is identical except for 5% -- this is often found where you have the same abstract but a few words are changed by the author while using a default abstract template like replacing "hypertension" for "diabetes". Another example is "Part I"/"Part II" abstracts. To review the original and duplicate references go to Project > Project Settings and select "Duplicates mode".
Another main issue is importing references from "Scopus" -- as they may have many of the same abstracts as Medline/PubMed -- but they have typed them in differently. Even the Journals are in a different format -- "Epidemiology and Medicine" vs "Epidemiology & Medicine". The way that I have dealt with this, at the start of the project, is to create a new Exclusion reason called "Excluded - Duplicate" and then sort by title or author -- and then manually go through all of the references looking at the title/author column to see if the same paper appears twice (note: I have found identical papers published in different journals). This method is rather fast -- but ensures you have removed any duplicates.