The GIVE project enters the home straight

In this bulletin, we provide an update on:

• Packaging and transportation (link in Dutch)

• Digitisation of newspapers, Flemish masterpieces and glass plates (link in Dutch)

• Metadata enrichment through artificial intelligence (link in Dutch)

• Results and knowledge-sharing (link in Dutch)

The GIVE newspaper project, Primeur, (in collaboration with Flanders Heritage Library) and the GIVE glass plate project are large-scale digitisation initiatives that involve processing a large volume of materials from various organisations. Digitising thousands of glass plates and newspapers on-site with the collection managers is simply not feasible, which means we have to transport them to our digitisation partners in bulk.

The first shipments to partners, GMS and Picturae, were sent in May last year, and the final deliveries are scheduled for this autumn. As we approach the completion of the digitisation process, we are of course also ensuring the safe return of the newspapers and glass plates to their custodians.

180,000 glass plates, 257 Flemish masterpieces and 630,000 newspaper pages – that’s a lot of material to digitise meticulously, page by page and piece by piece. This summer, we have again been very busy:

Since January, our digitisation partner has been providing a digital version for each newspaper page. And, despite the high piles of newspapers to work through, they have been making good progress: we have already completed three-quarters of the total amount. 

Here’s how Picturae is patiently capturing each page:

Can’t see the video? Please check your cookie settings and allow us to show you this content. You can change your cookie settings on this page. Click on ‘Change your consent’ above the table and select ‘Preferences’ and ‘Statistics’.

Nearly 260 masterpieces, including rare and invaluable works of art, manuscripts, drawings and sculptures, will have passed through our hands by the end of the project. We only have eight more items to digitise! Everyone works hard during a day of digitisation at St. Leonard’s Church in Zoutleeuw. Come and have a look with us! (link in Dutch)

In January of this year, we embarked on the task of digitising over 180,000 glass plates, and have now successfully created digital versions of all glass negatives, positives and lantern slides. The digitisation process took place primarily in the Netherlands, with two-thirds of the glass plates being digitised there. The remaining plates were digitised at the Boekentoren (Ghent University Library’s ‘Book Tower’) and FOMU (Fashion Museum of Antwerp). In August, we relocated our mobile digitisation studio to the latter location. Join us for a look behind the scenes.

In the GIVE project, we are making a vast amount of Flemish heritage more accessible, searchable and reusable. In addition to digitising new works, we are also working with materials already present in the meemoo archive system: thousands of hours of audio and video from previous digitisation initiatives and digital intake. These are treasure chests of information that are difficult to search due to a lack of descriptive data. We are therefore integrating three methods of artificial intelligence (AI) into our systems to enrich them. So where do we currently stand? And why is the end of September such a pivotal moment?

We are using the Speechmatics Software-as-a-Service (SaaS) application to convert the audio files in our meemoo archive system into transcriptions. Applying this ready-made model to Flemish – and mostly Dutch-speaking – archival content is not straightforward, however. We therefore dedicated the spring months to integration, tuning and testing. And it paid off! Between June and July, the system processed all 160,000 hours of audio. The results:

• Approximately two-thirds successfully transcribed;

• One-sixth of the files do not contain any speech;

• For the remaining files, Speechmatics could not determine the spoken language with sufficient certainty. We will be working on a good approach for this batch over the coming months.

The decision has been made! After making some thorough comparisons, we have selected the TextRazor SaaS application. This tool allows us to extract relevant place names, personal names, organisation names and other entities from the initial transcriptions – additions that make searching for archival content even more accessible. We pressed the start button in August and are now halfway through processing. We will know the results by the end of September.

Bonus: TextRazor not only performs Named Entity Recognition (NER), but also links the identified entities to publicly available authorities, such as Wikidata.

What if you could find out which videos a particular public figure appears in with just one search query? We are making this possible with the facial recognition component – building our own application for it, which will remain usable even after the project ends.

Want to see the results? Following a thorough quality control process, we will gradually make the digitised newspapers, masterpieces and glass plates accessible on our own platforms, and on our participating partners’ platforms if possible. We’re also preparing the generated metadata so that it can be re-used easily and efficiently. This requires some work, however, which is why we will be highlighting some beautiful gems and sharing the knowledge we have gained in the meantime. After all, isn’t the strength of this impressive collaboration in the details? More on this soon!

Stay informed about everything GIVE:

The GIVE projects have been made possible thanks to support from the European Regional Development Fund and are part of the Flemish government’s Resilience Recovery Plan.

Sources visual digitisation:

• Lantern slide from the collection of the Toy Museum

• The Golden Book of the Chapel of Venerable, volume I, Chapel of the Most Blessed Sacrament in the Cathedral of Antwerp, 1631/1840

• Newspaper De Werker, 1932, from the collection of Amsab-ISG