The time machine exists. But it doesn't look like the sparkling DeLorean from 'Back to the Future'. His pieces are scanners and computers. These are the work tools of Joana Maria Pujades Y Alícia Fornés, two researchers who have managed to penetrate the past of Barcelona as nobody had done before.
The two scientists have analyzed with artificial intelligence hundreds of thousands of marriage certificates, registrations and other documents, written over the last five centuries in Barcelona and surroundings. The results have broken topics. For example, the Catalan rule of 'hereu' was not as strong as it seems. Or the industrial revolution did not increase inequality much: rather it changed its nature.
A social network with the past
Part of his work has turned to a platform that looks a bit like Facebook. The system allows access to individuals who, surprisingly, lived decades ago, when they could not even imagine that they would end up in a social network. Big data conjugated in the past.
Researchers have experience in all kinds of documents. For example, Fornés has worked even with mechanical piano rolls. It has converted more than 3,500 of them into files that can be heard on the website of the National Library.
Now, specialists intend to use this experience to process the most crazy information: from photographs and pictures to maps and staves. All this should converge on the Catalonia Time Machine: a digital platform that allows the most complete and realistic immersion possible in Catalonia's past.
Work part It appears in a kind of Facebook that allows access to locals from other times
This is the local version of a string of similar 'time machines', which are being developed in various parts of Europe, from Venice to Amsterdam. «We want to automate the reading of population documents, such as betrothal and records. If a document says: 'Pere Puig marries Maria Vila in the parish of Pi', our system must understand what is the name of the husband, wife, where they got married, etc. ”, explains Pujades, who is head of the area of Historical Demography at the Center d'Estudis Demogràfics (CED), in Bellaterra.
In 2010, under the guidance of Anna Cabré, Pujades set to work with 600,000 marriage certificates registered in the diocese of Barcelona between 1451 and 1905. This is a unique record, because the diocese had the privilege of charging a proportional income tax for the bride and groom, which «Allows to study inequality for five centuries», explains Pujades. The processing of the documents revealed that lInequality in the city was also high before industrialization. What happened with the industrial revolution is that it changed cause. Before, those who widened inequality were the nobles, who accumulated wealth. Then there were the workers, increasingly poor.
The rule of ‘hereu’
"With the traditional method, this project would have lasted until my retirement," says Pujades. «We have put technology into something very traditional, which is to collect the sources. We have modernized and accelerated it, ”he explains. The names of the parents also appear on the marriage certificates. Pujades and his collaborators took advantage of it to draw family trees and study how wealth was spread throughout the generations.
According to the rule of the 'hereu', the firstborn keeps all the heritage. So the researchers expected the other brothers and sisters to be impoverished. However, marriage taxes reveal that does not happen. «Although they do not inherit, they remain in the same social class. It is probably the result of cooperative strategies. For example, allow the brother's flock to graze in the lands of the 'hereu' », explains Pujades.
Building the Big Data of the past is a path of challenges. "The paper of the manuscript can be degraded, the spelling varies a lot, the vocabulary is old & mldr;", explains Fornés, a researcher at the Center for Visió per Computador (CVC), also in Bellaterra.
A discovery is that inequality did not increase with the industrial revolution, but rather changed its nature
Fornés addresses these challenges through 'deep learning', the same principle that Google uses in its services. "We want computers to learn to read," he explains. In practice, the computer is trained to read familiar words; then, to read all the words written by the same hand; later, similar hands, and finally taught to interpret documents of different types.
To validate and reinforce machine learning, Pujades and Fornés have devised an unusual method: a mobile video game. Anyone can download their "word hunter." The dynamics of the game forces the player to identify if, among five handwritten terms that the computer has classified as equal, there are some actually different.
"The difference with traditional methods is in the amount of information that is handled, but this ends up causing a qualitative difference," says Joan Anton Barceló, professor of the Department of Prehistory of the Autonomous University of Barcelona, and one of the standard bearers of the digital humanities in Spain. "In the case of Barcelona, it is already beginning to be seen that some of the great historical reconstructions do not add up," says the researcher, who is not involved in the work of Pujades and Fornés.
"The algorithms do not recognize social connotations or mistakes, ”says Professor Tim Brennan
Barceló would like to go even further. «The great revolution is artificial societies: create a simulation of a civilization on a computer. What was previously described with a text is now an algorithm that can be checked with empirical data, ”he says. Barceló hates research loaded with ideological frameworks and trusts that digital humanities reduce arbitrariness.
Algorithms and criticism
This position generates rejection in part of the humanist community. «There is a lot of mystification about the power of algorithms. In reality, they are as good at recognizing patterns as those who program them. They extract only what you want them to extract, ”says Tim Brennan, professor of Humanities at the University of Minnesota and critic of the digital humanities.
"Computers cannot recognize social connotations, irony, mistakes," he adds. Brennan believes that the growth of digital humanities is encouraged by the willingness of the departments to raise funds from technology companies and to do without workers.
"We want computers learn to read, ”explains researcher Alicia Fornés
«The most important thing is that this approach displaces critical thinking. It tends to ignore that there is no position outside an ideological framework, ”he says. Even so, Brennan concedes that computers can be useful for processing large amounts of material if combined with human control.
In this, it coincides with Fornés. «Artificial intelligence will not replace humanists. But it will allow you to speed up your work and suggest patterns that you should interpret. We believe it is essential to keep humans in the process, ”he says.
The researcher remembers the case of a document with a big stain on some words. The computer could not read them. Then came a paleographer who, looking at the time, the style and the parish to which the document belonged, sentenced that there was no doubt. The words under the stain were "wool weaver."
. (tagsToTranslate) Barcelona (t) history (t) Big data (t) artificial intelligence