Why does neural machine translation sometimes generate translations with high BLEU scores but lack coherence?

💡 Explanation

The attention mechanism focuses on individual word pairs, maximizing short-term dependencies, but fails to capture long-range contextual information which leads to incoherence. Therefore, high BLEU scores can be achieved due to accurate local translations, rather than reflecting overall text quality as affected by statistical model overfitting.

• Why does adoption of a non-native accent increase perceived prestige in international call centers?
• Why does neural coding in the auditory cortex exhibit sparseness, where only a small subset of neurons respond strongly to a given sound, rather than a broad activation?
• Why does the phonological structure in sign language permit greater morphological complexity than spoken languages?
• An engineer localizing an educational video game for younger children chooses to simplify cultural references and gameplay mechanics. Which outcome is most likely from this design choice?
• Within a corpus linguistics project analyzing historical legal texts, which risk increases as annotation scope broadens without adjusting for inter-annotator agreement?
• Why does text using a monospaced font still require subtle kerning adjustments for optimal legibility in a printed book, despite its inherent fixed-width character cells?