The TU Delft Library Main Hall’s bookshelves contain 25,000 volumes. In the basement archives, there are tens of thousands more books. Digital collections will soon match and exceed these numbers. There are over 60,000 documents in our institutional repository and the number grows exponentially.
Referencing and searching documents has long been the business of the library. Library science has a long history with traditions and methods that are being reimagined in the digital landscape. Libraries across the globe are exploring how digital technologies can visualize and activate collections.
With the first Oracle prototype, the TU Delft Library and RNDR look to the future of data visualization for institutional repositories. The custom-coded project uses machine-learning processes such as graph embeddings, text embeddings and autoencoders to create multi-dimensional (of up to 500 dimensions) maps out of the large dataset of TU Delft student theses and dissertations.