Include a small number of segments that follow the current segment when sending text to MT. Provides extra context for the model to reduce incomplete or ambiguous translations, while keeping token count manageable. Balances accuracy with performance constraints.