A young computer scientist from Santiago devises a system that translates meteorological data into natural language.

A surprising picture: the Obradoiro square cloudy after the rain.

An application to know the weather that will be in Galicia may sound like a joke, because the cliché is well established. But as the Ourense writer Juan Tallón wrote, “any countryman knows that it doesn’t rain in Galicia and that is why important things are always done in the open air: the barbecue, the carnival, the processions, the fair, the proclamations, the fires of the Apostle, the Santa Compaña … “

No matter what it sounds like, the approach to weather forecasting that Alejandro Ramos Soto, PhD in Computer Science from the University of Santiago de Compostela, has made with his Galiweather application is a huge advance, that of eliminating weathermen and any other intermediary. This program extracts meteorological data and translates it directly into natural language, offering citizens a prediction for each of the 314 councils gallegos.

“As far as I know, there is nothing like this in Spain or in Spanish,” the Galician computer scientist told this newspaper. “When we developed Galiweather we did it motivated by research, although that ended up being a real application that we gave free to MeteoGalicia ”, the community’s meteorological service.

How does it work?

The software extracts relevant meteorological information from the data for the next four days: wind intensity and direction, maximum or minimum temperatures … “and on that a series of techniques are applied that extract the relevant information in a non-textual way, but with a kind of proto-language that, using fuzzy logic techniques, gives us a degree of veracity on the information that we are gathering ”.

Fuzzy or fuzzy logic is a way of converting numerical values ​​into something that resembles what humans do in the real world. For example, if we ask someone if it will be cold in the afternoon, nobody usually answers “there will be an 84% chance that the temperatures will drop below 12 degrees”, but rather “go, take a jacket”. This interpretation work is traditionally performed by meteorologists, but the Ramos system has managed to emulate it with some success. “Depending on the degree of truth that the system obtains, we decide how we verbalize it, for example: most of the next few days it will be cloudy or clear: we use linguistic terms instead of concrete percentages.”

Recently, this finding has earned the young computer science doctor one of the researchers of the year awards from the Spanish Scientific Information Society and the BBVA Foundation.

“There is very great potential in the use of fuzzy logic techniques to natural language “, explains Ramos, who contemplates expanding the application of Galicia to other areas of Spain,” as well as looking for other applications for the system in areas such as health or business intelligence“.