Artificial intelligence trudges through Italian museums

At the Veneranda Biblioteca Ambrosiana, the linguistic aspect has been intertwined with the cultural one by making accessible a digitised corpus of 250 ancient manuscripts in Arabic from different eras. But AI does not only have a translation function from Arabic, it makes very complex manuscripts readable: 'Arabic already presents the greatest difficulty in linguistic terms. Then the material is manuscript, with handwriting that is not encoded in a standard way, as is the case in printing. And then the layout of the writing occupies the entire page, in horizontal, vertical and oblique dimensions. It seemed to us the best possible terrain for experimenting with the technologies we implemented,' explains Fabio Cusimano, Head of Digital Asset Management Cataloguing at the Veneranda Biblioteca Ambrosiana.

AI Htr Handwritten text recognition technology was therefore applied. "We developed the technology with the aim of recognising the page layout," Cusimano points out. Now the entire corpus (another 1,500 manuscripts) will be submitted to AI and made available to everyone on the Milan institution's website, opening up interesting research perspectives: "Alongside our institution's preservation objectives, which are enhanced with the digital replacement copy, we are offering the best tools to the scholarly community and the wider public," Cusimano explains. It is not so much a matter of transcribing or translating the text but of offering a platform that interprets the content and generates a huge amount of text that can be interrogated by users. Who can then contribute to improve and refine the content.

This cultural project would obviously not have been possible without the digitisation that started in 2018. However, the Polytechnic's research reveals a very gloomy general trend in the Italian situation: digitisation of collections in Italian museums has only marginally increased in recent years, and 26% of institutions state that they have not yet started any activities of this kind, while 23% of museums still do not collect visitor data. In many cases, an integrated data collection, interoperability and exploitation strategy is still lacking, which is also likely to severely limit the future effectiveness of artificial intelligence applications.

"In ten years there have been many changes, but we must recognise that so much of the cultural ecosystem has remained very close to what it was ten years ago, despite the resources of the NRP that could have induced a leap forward," Lorenzini comments.

A second megatrend noted by scholars is that of AI enhancing the work of operators, which finds further and more widespread applications in programming, audience analysis, collection curation, predictive maintenance, fundraising and decision support. Finally, another megatrend concerns the development of collaborative data sharing practices through shared data ecosystems, in which cultural institutions and organisations work together to exchange information in a structured, secure and interoperable way. 'These practices,' reads the report, 'will enable integrated management of cultural organisations and networks, supporting joint analysis, cross-institutional services and cultural policies based on collective evidence, including from other industries'.