Curating Our Digital Cultural Heritage: What’s Next?

Having shared previous posts on the motivations behind our OSC , as well as more detail on our digital tools and processes, we will now reflect on our experiences from this project and think about how they might inform future work. Keep reading for information about how you might be able to get involved in our work, as well as a sneak preview of our next OSC.

To recap, our previous post ended by identifying three challenges from our OSC that might be addressed in future projects:

• Enhancement
• Scalability
• Interoperability

We’ll now go through each of these in turn, to discuss the issues in more detail and suggest possible solutions.

Enhancement: how might we ensure a richer data export from Recogito Studio to provide more informative visualisations?

While we were able to export geographic data from the Recogito Studio annotation platform and visualise it as a map using the Peripleo GitHub resources, we found that this data was lacking in the key information contained in our non-geographic annotations. For example, when someone clicks on a map marker to find out more about the object to which it relates, it would be helpful to provide more information about that object’s itinerary (its journey through space and time), rather than just its identification number and catalogue link. This might be achieved through further development work on Recogito Studio, to give users the option to export this enhanced dataset. Alternatively, development could take place outside Recogito Studio to produce a basic tool that might convert the full annotation export CSV (as opposed to the GeoJSON export that specifically provides geographical data) into a format that can more easily be uploaded into Peripleo for visualisation as an online map. Our colleague Daniel Pett, who also worked on this OSC, has used Historic England’s Heritage at Risk data to produce an enhanced Peripleo visualisation - if this type of data could be exported from Recogito Studio and plugged straight into Peripleo, that could be extremely powerful.

Image: geographic data can be exported from the Recogito Studio annotation platform and visualised as a map

Scalability: how might our processes be automated for use on larger datasets?

The work that we have done for our pilot study relied solely on manual annotation, i.e. a human user had to individually highlight each piece of text, tag it with the relevant keywords, and (for places) align each reference to the relevant record in a global authority file, like Wikidata. This process works perfectly well for our small dataset of catalogue records for nine museum objects; it would, however, be impractical for scaling up to annotate larger collections. To address this, we would need to look at ways in which annotation might be automated, by investigating techniques such as Named Entity Recognition (NER) or harnessing the power of Artificial Intelligence (AI) through the application of Large Language Models (LLM). Some NER functionality was available in Recogito Studio’s predecessor, by automatically tagging places, people and events, as shown in the screenshot below. Before similar functionality can be implemented in Recogito Studio, significant work must first be done to identify the most appropriate solution for a cultural heritage context, while also examining the potential risks and ethical issues of automating annotation using these methods. Then, once a solution has been identified, the required technical developments might take place within or separately from Recogito Studio, or a combination of the two.

Image: Named Entity Recognition (NER) functionality can automatically tag places, people and events.

Interoperability: how might we use our exports from Recogito Studio to produce data that can be more readily linked to other resources?

One of the advantages of the original Recogito platform is that it could export semantic annotations as Linked Open Data, i.e. data structured using standards that present the different pieces of information (say, places and their alignment to global authority files), in a way that can be understood by a computer. Structuring data using Linked Open Data principles therefore enables connections to be made with other datasets and digital resources. Because the new Recogito Studio was designed first and foremost as an easy-to-use collaborative annotation platform, functionality to export data that adheres to these standards is not built in, though, due to the development of an extension mechanism supported by this OSC, it is certainly possible. Looking more specifically at data structured using the data model and annotation protocol described in our previous blog post, it would be possible to develop a tool (or an easy-to-follow workflow that uses one or more existing tools) that would convert a Recogito Studio export CSV to a Linked Open Data format that aligns with this data model. In doing so, this would mean that different collections datasets can more easily be brought together based on their common features, reveal new insights and connections, and increase the potential for digital storytelling.

So, what next? We are fortunate to be able to keep up the momentum of this project by starting a new OSC in March. While we will continue our evaluation of Recogito Studio for annotating cultural heritage data, our attention this time turns to a community organisation. Volunteers from the Campsbourne Community Collective will use the platform, alongside the Memory Mapper software, to curate their own digital heritage to make it more discoverable to a wider public. We’re really excited about this project and look forward to sharing the results here in due course.

If our current OSC has captured your interest and you’d like to be more involved in working with historical data, do have a look at the Pelagios Network – it’s a helpful and friendly knowledge sharing community and is free to join. Alternatively, if you’d just like to find out a bit more about this particular project, please visit our Challenge webpage.