Usually, artificial intelligence and big data suggest futuristic applications such as self-driving cars or advanced financial risk assessment, but this technology is also proving its potential in the analysis of the distant past. Thus, the digitalization of extensive historical archives is opening the door to an in-depth review of massive amounts of documents that was previously unfeasible for researchers. Undoubtedly, one of the most extensive archives of its kind is the General Archive of the Indies (AGI, by its initials in Spanish), hosted by the Spanish city of Seville, which keeps a meticulous account of the relationship between Spain and the American colonies, as well as the travels of its discoverers. Eighty million handwritten pages. Those numbers are naturally out of reach for conventional analysis and only 10 percent of the documents have been studied so far. Artificial intelligence, however, has soon chewed through them and provided some incredible insights that could even change history as we know it.
This Artificial Intelligence project begins with Enrique Vida, a physicist who works for the Universidad Politécnica de Valencia (Science University of Valencia), came up with the idea of developing a software able to identify words in ancient manuscripts. In addition to different handwriting styles, the main challenge is the variations in orthography throughout the centuries. That was one of the main reasons to take historians on board. Vidal, who heads the Pattern Recognition and Human Language Technology (PRHLT) research center, reached out to Carlos Alonso, a paleographer that provided some of the clues to develop “Carabela” (caravel), the new Artificial Intelligence software that they recently announced.
Following the initial analysis of 514 documents as a sample of handwriting and orthography variations, Carabela processed 60,000 digitalized records from the AGI plus 90,000 from the Archivo Histórico Provincial de Cádiz (Regional Archive of Cadiz). The software already offers an 85% accuracy for keywords that are searched regardless of their spelling and taking into account other terms within the same semantic field. Thus, when they entered the term “boat” other words like “galleon” or “ship” were also parsed thanks to Artificial Intelligence. This specific search revealed 150 unheard-of shipwrecks. But the most striking discoveries were yet to come. Next they decided to enter the term “austral,” a search that produced unexpected results. Out of the blue, an old text from a Jesuit priest pointed out the location of Australia and the surrounding islands, as well as mentioning that seafarers Vaes de Torres and Quirós had discovered them a century before. Their journey is dated in 1606, more than half a century earlier than James Cook’s claim of the continent for the British empire.
AI reveals a new geoglyph in Nazca
For years, the giant figures drawn over the Nazca desert have sparked the interest of researchers. Their vast scale only allows them to be seen from the sky as if their unknown authors would have drawn them to honor their gods. Now, researchers at IBM jointly with the University of Yamagata in Japan have developed an Artificial Intelligence software that searches for glyphs from satellite images. That is how they have located a hitherto unknown humanoid figure, later confirmed by in-person visits to the site. The next step will be to improve the efficiency of the AI software by using laser mapping (LIDAR), a technology that has already located lost cities in the Amazon rainforest.