New AI-Based System Developed to Detect and Assess Noisy Activities from Social Networks Data

Scientists from Universidad Politécnica de Madrid (UPM) have designed a system of text analysis that, applied to comments posted in social media, is able to automatically pull up grievances on noise pollution and categorize them based on their origin. This system integrates artificial intelligence (machine learning) with different methods of language analysis.

Furthermore, this system can predict the appearance of noisy events, which can help city managers to prepare early interventions to prevent disturbances and health-related issues for citizens. This research has been developed in partnership with Télécom Paristech.

It is estimated, in Europe, that 25% of the population experiences high noise levels, which can increase health hazards. This causes public health problems and diminishes the quality of life, particularly in urban areas which are related to the lack of rest and stress because of an unsettling situation.

Traditional surveys have been used to understand more about citizen insight regarding a noisy urban environment, but they have the disadvantage of high cost deriving from its development and implementation, the narrow number of participants, and the length of the surveying campaign.

Also, this system is not swift when perceiving problems or precise noisy events. In the last few years, new systems of online citizen contribution have been set up which allow a proper interaction with local managers, but they are not generally used by the people as they probably do not feel comfortable using this system.

Social media users give their opinion and feelings about various topics: politics, products, TV, and of course, environment, including noise pollution.

For years companies have been applying machine learning and natural language processing techniques to find out the opinion of clients about their products and services in social media in order to improve sales. However, this technological trend has not been applied in city management, missing an opportunity of a channel used by thousands of citizens such as social media which can provide real-time data about issues in a city.

Luis Gascó, Scientist, Instrumentation and Applied Acoustics (I2A2), UPM.

Like this, the project research group has created a text analysis system that can automatically detect complaints on noise pollution and categorize them based on their origin.

To realize this, they used the newest methods in artificial intelligence, such as machine learning, and varied methods of language analysis. Furthermore, scientists have engineered a forecast system using statistical methods that permit them to discover the appearance of a disturbing noise event from the number of complaints and detailed words.

The application of the system designed by UPM scientists is not only restricted to the field of noise pollution, the scientists say “this system could be used to detect other types of problems, from damages in the urban furniture to the opinion of the citizenship about changes in urban planning, for instance, the semi-pedestrianization of Gran Vía in Madrid.”

Currently, the scientists together with their French colleagues from Télécom ParisTech are seeking new partners, mostly companies dealing with city management and transport infrastructure, to take part in a project of technological transfer where this technology can be tested in vivo.