Identify segments within a single document that are very similar in source text but have different translations. Group them together to allow editors to standardize wording and maintain internal consistency. Uses normalized source text and trigram similarity to efficiently find near-duplicates.