In the context of cultural heritage dissemination, machine translation tools can facilitate access to multilingual content. However, their probabilistic nature raises questions about their reliability. This study presents preliminary findings from a machine translation test conducted within the UniVOCIttà project. A qualitative error analysis was carried out on a sample of English and French excerpts describing Bologna’s main art gallery, the Pinacoteca Nazionale, translated into Italian using two neural machine translation systems (Google Translate, DeepL) and two LLM-based chatbots (ChatGPT 4o, Claude Sonnet 4). The inaccuracies observed in the generated outputs suggest that these tools are not yet fully reliable for specialized translation tasks. These findings underscore the need for collaboration between cultural institutions and language professionals to ensure the dissemination of accurate multilingual content about cultural heritage.
Gramellini, R. (2026). Machine Translation for the Multilingual Dissemination of Cultural Heritage: A Case Study from the UniVOCIttà Project. Bologna : Alma Mater Studiorum Università di Bologna.
Machine Translation for the Multilingual Dissemination of Cultural Heritage: A Case Study from the UniVOCIttà Project
Rita Gramellini
2026
Abstract
In the context of cultural heritage dissemination, machine translation tools can facilitate access to multilingual content. However, their probabilistic nature raises questions about their reliability. This study presents preliminary findings from a machine translation test conducted within the UniVOCIttà project. A qualitative error analysis was carried out on a sample of English and French excerpts describing Bologna’s main art gallery, the Pinacoteca Nazionale, translated into Italian using two neural machine translation systems (Google Translate, DeepL) and two LLM-based chatbots (ChatGPT 4o, Claude Sonnet 4). The inaccuracies observed in the generated outputs suggest that these tools are not yet fully reliable for specialized translation tasks. These findings underscore the need for collaboration between cultural institutions and language professionals to ensure the dissemination of accurate multilingual content about cultural heritage.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.



