14.09.17 - Sofia Ares Oliveira works at the Digital Humanities Laboratory and focuses mainly on image processing. She develops computer vision and machine learning tools enabling the extraction of meaningful information from digital images.

After studying Electrical Engineering at EPFL where she specialised in signal processing during her master, Sofia joined the Digital Humanities Laboratory. Here she deals with digitised archives mainly from Venice. The Venetian Archive has a very large collection of digitised documents (and even larger not-yet-digitised collections), so large that only a small percentage has been viewed once. The benefit of an automatic extraction and processing of the information contained in these documents is huge, as it eases the discovery of new elements and enables a faster search among the documents.

Handwritten text recognition

In the context of the DHLAB collaboration on the european project READ (Recognition and Enrichment of Archival Documents), part of Sofia's research focuses on handwritten text recognition, where she develops tools enabling an automatic reading and transcription of text documents. These kind of systems use neural networks and are trained by being shown many examples.

Sofia also works on ancient Venetian cadaster maps. Here she develops tools that allow the extraction of the shapes of the plots of land and that automatically labels them. These elements are then linked to the cadaster registers and used to study the history of the city and its evolution.

Source:Institute of Digital Humanities