A machine learning engineer is designing a system that uses a specific type of attention mechanism. In this mechanism, the calculation for any given word in a sentence can only incorporate information from that word and all the words that came before it; it is explicitly prevented from using information from any subsequent words. The engineer is considering using this system for two distinct tasks:

Task A: A real-time translation service that begins translating a sentence as the user is typing it. Task B: A sentiment analysis tool that classifies a completed movie review as positive or negative after the entire review has been submitted.

Analyze the suitability of this specific attention mechanism for each task. For which task is it well-suited, and for which is it poorly-suited? Justify your reasoning for both cases.