Why does machine translation struggle with deictic expressions across languages more than with other word classes?

💡 Explanation

Machine translation has more difficulty with deixis due to the reliance on shared contextual knowledge between the speaker and listener, because the contextual information may not be present, or different in the target language, therefore, the translation becomes ambiguous. Rather than universal semantic roles, the contextual dependency makes accurate translation complex.

• Why does a young vervet monkey sometimes issue an 'eagle' alarm call upon seeing a harmless bird, unlike its experienced elders?
• Why does the channel capacity of a quantum communication system using entangled photons degrade under decoherence?
• A museum tour guide points to a specific artifact while saying "This one is particularly significant." Which mechanism is most amplified by this indexical reference?
• A remote island community experiences sustained contact between English-speaking traders and speakers of a local Austronesian language; if a stable creole emerges, which consequence is most likely?
• A speech recognition system misinterprets 'dust' as 'gust'. Which mechanism explains why this substitution occurs in acoustic phonetics?
• During online question-answering, why does providing 'given' information already known to the system affect a user's interpretation of the response?