In a June 27, 2024 paper, researchers from Unbabel and Instituto de Telecomunicações introduced xTOWER, a large language model (LLM) designed to generate “high-quality” explanations for translation errors and use them to suggest improved translations.

The researchers explained that machine translation (MT) systems, despite their strong performance, often produce translations with errors. “Understanding these errors can potentially help improve the translation quality and user experience,” they said. 

Built on top of TOWERBASE — an LLM designed, trained, and optimized for MT-related tasks —, xTOWER offers detailed, human-readable explanations for translation errors and suggests corrections based on this analysis.

Specifically, the process involves inputting a source text and its translation into xCOMET, which annotates the translation with error spans and assigns a quality score. The complete input (i.e., the source text and its translation), the annotated translation, and the quality score are then passed to xTOWER, which generates explanations for each error span and proposes a new corrected translation based on these explanations.

Source: https://slator.com/

Full article: https://slator.com/unbabels-new-xtower-llm-explains-translation-errors-and-suggests-how-to-fix-them/