Artificial Intelligence is going to tell you how good the air quality is in your town

It is called ICA2TEXT, and uses techniques based on Artificial Intelligence (AI) and Natural Language Technologies to automatically generate textual descriptions in Galician and Spanish about air quality data in the Autonomous Community of Galicia. The tool, recently deployed as a public service on the website of the regional meteorological service, was developed jointly by the Air Quality Department of Meteogalicia (belonging to the Directorate General for Environmental Quality, Sustainability and Climate Change), and the Singular Centre for Research in Intelligent Technologies of the University of Santiago de Compostela (CiTIUS).

ICA2TEXT accompanies the graphs and numerical-symbolic data displayed on the web with automatically generated textual descriptions, which help citizens to have a better understanding of the information provided. The aim of this application is to describe these data in an understandable way, just as an expert would. The project is a step forward in the line of collaboration that the two institutions have been maintaining over the last few years, as a result of which similar initiatives have crystallised, such as GaliWeather, a pioneering tool in the world that, since 2015, 'translates' every day into natural language the weather forecast data of the more than 300 Galician city councils.

Artificial intelligence in the service of meteorology

The Air Quality Index (AQI) is a parameter that is automatically calculated with the data emitted in real time by the air quality measurement stations spread throughout Galicia; until now, the result of this information collection was shown to users on the Meteogalicia website through a simple visualisation mechanism, based on a colour code.

In this context, ICA2TEXT presents a very novel proposal, based mainly on two techniques that are currently widely used by the scientific community of experts in artificial intelligence: Natural Language Generation (NLG) and Fuzzy Logic. "This new application is mainly composed of two main tasks: on the one hand, it processes the data to generate coherent linguistic descriptions from them,' explains the researcher in charge of the project at CiTIUS, Alberto Bugarín; 'on the other hand, it is capable of converting the descriptions generated into texts that people can understand. An innovative system, framed in the area of Data-To-Text (D2T, 'Data to Text') that uses techniques based on Artificial Intelligence (AI) and natural language technologies to automatically generate textual descriptions.

The work began in 2020 within the framework of a Final Degree Project (TFG) of the USC's Degree in Computer Engineering and has been presented* at the XXXVIII Congress of the Spanish Society for Natural Language Processing *(SEPLN). It was also published in Applied Soft Computing, one of the high impact scientific journals in Artificial Intelligence.

"As far as we know, no meteorological agency in the world currently has an application similar to ICA2TEXT for the automatic generation of natural language descriptions of the Air Quality Index," says Bugarín. "All this gives an idea of the projection and importance of this result, which encourages us to continue improving our collaboration with Meteogalicia in the application of intelligent technologies and natural language to the field of meteorology", he concludes.