Artificial Intelligence and Heritage
AI cannot only be reduced to chatbots and intelligent digital assistants. It can be applied in many more fields than we can imagine, while steering clear of autonomous AI we see in Science fiction.
In this article we invite you to discover a few AI projects applied to written heritage, which bring together librarians, developers, linguists as well as many other specialists.
The NewsEye project
Funded by the European Union’s Horizon 2020 research and innovation programme, it involves six universities and two European libraries. It aims to improve and modernize Humanities research thanks to digitized historical newspapers.
AI has been introduced in the field of documentation science through optical character recognition (OCR). This technology developed in the United States allows for the recognition of printed, typed or handwritten text. NewsEye’s researchers use a trained OCR technology capable of recognizing text as well as article segments even when they are damaged. Besides being able to do a full text search, the user is also aided by a personal research assistant, which allows the user to analyze and sort through articles in order to compare them and conduct statistical trend analyses. A considerable advantage: the tool is multilingual and can handle searches in various languages (implemented beforehand). The Newseye platform is already available (you only need to create an account to access it).
Snoop
Snoop is the image indexing and visual search engine developed by the National Institute for Research in Digital Science and Technology (Inria) and The National Audiovisual Institute (Ina).
The tool is being tested at the BnF (National Library of France) and enriched with the digitized collections from Gallica. It can find visual content using a key word or an image chosen by the user. The user can also create personal collections and therefore contribute to the training of the AI by validating or not the suggested images.
The BnF already uses it for cultural mediation, for example in this blogpost on maritime inventions.
Discover GallicaSnoop !
« LITTE_BOT »
A robot who can interact with literary works, characters or their author, LITTE_BOT is a chatbot bringing Dom Juan to life. Imagined by Rocio Berenger, it was available at the BnF (Richelieu site) during the exhibition which took place in September 2022, celebrating the 400th anniversary of Molière’s birth. This chatbot was developed using the semantic analysis of a corpus of 400 17th century plays.
By isolating the lines of dialogue by theme, it attempts to answer coherently to its human conversation partner. Initially, the project was supposed to also allow this virtual Dom Juan to learn how to invent new phrases, however this ambition was discarded due to time constraints. Would another character be able to exchange with us on his own terms?
This project, whose scientific aspect cannot be denied, reveals an entertaining, artistic and linguistic side to how we use AI with our cultural heritage.
Pierre Brissart, frontispice for the
OEuvres posthumes de Monsieur de Molière edition, t. VII, 1682.
You don’t get the chance to carry a conversation in 17th century French every day.
The database of the Répertoire des écritures manuscrites du département de la Musique (REMDM)
The database of the Répertoire des écritures manuscrites du département de la Musique (REMDM – Directory of Handwritings of the Music Department) and its automatic image mining tool developed with the IT labs L3i (La Rochelle University) and IRISA (CNRS) will soon allow music amateurs or professionals to find out whether a music sheet was handwritten by its composer or by a copyist.
For now, this database is still in the development phase and mobilizes librarians from the BnF’s Music Department and musicologists in order to correct and enrich handwritten music sheet holdings and to identify the handwriting of music writers, composers or copyists. Thanks to the automatic image mining tool, it will be possible eventually to identify a writer and to analyze and compare different handwriting.
Autograph manuscript by Joseph-Hector Fiocco,
composer (18th C.)
Conclusion
In the past 20 years, the use of AI applied to heritage hasn’t stopped evolving. Thanks to AI, data that until now was difficult to use is now available to the general public. A genealogy enthusiast or a humanities researcher has nowadays much easier access to the content of documents, which are increasingly likely to provide them with quality data thanks to AI. Meanwhile said documents are manually handled less often and their conservation is therefore improved without having to restrict the access to them. It’s a win-win situation for both the document and the reader.
Some people might wonder if librarians would not be replaced by AI in the near future. The answer, supported by many professionals, is no. The applications presented in this article highlight the fact that human action is needed in order to develop and broaden the knowledge of the AI. Its framework goes beyond human understanding when it compiles multiple tasks in a few milliseconds, however it is limited by the AI’s operating data.
For librarians, the dawn of AI involves only a review of practices, just like IT and internet changed library management systems.
AI evolution is in good hands with librarians.
Written by CLELIA ROOS