Researchers develop better translation for Coptic, a dead language, by combining dictionary lookups with grammar rules

A research team published a new method for translating Coptic (an ancient Egyptian language with few speakers and little digital text) into English by combining word-by-word dictionary matches with grammatical structure analysis from a tool called Universal Dependencies.

Unlike prior approaches that relied only on word dictionaries, this method adds explicit grammatical information—showing the AI how sentences are constructed—which together achieved measurably better translation accuracy across different AI model sizes.

This matters for linguists and historians studying ancient texts: previously, translating low-resource languages like Coptic was nearly impossible with standard AI tools because there wasn't enough training data. This technique shows how to bridge that gap by teaching the AI about sentence structure, making rare-language translation more practical.