Why has high-quality automatic machine translation been difficult to achieve?

Created from Britannica articles using AI. AI can make mistakes, so verify using Britannica articles.

Ferdinand de Saussure seated in a formal pose, wearing a dark suit with a vest and sporting a thick mustache, circa 1900. — Linguistics Much of the foundation of 20th-century sociolinguistics was created by the Swiss linguist Ferdinand de Saussure (1857–1913).

High-quality automatic machine translation remains challenging because languages are deeply embedded within cultures, which differ significantly from one another.1

The primary difficulty lies in the lack of one-to-one lexical equivalence between languages. Words often carry unique overtones, historical associations, and cultural nuances that are not shared by their nearest counterparts in other languages. A literal, word-for-word translation frequently fails to capture the intended meaning because it ignores these complex, interwoven layers of context.1

These challenges are particularly pronounced in literary works and poetry, where the author exploits specific linguistic resources—such as word order, grammatical constructions, and phonological features like metre or rhyme—that are unique to the source language. While technical prose is easier to translate due to existing international scientific standards and higher degrees of lexical correspondence, capturing the stylistic and semantic depth of natural language remains a complex task for machines.1