Beyond sound level monitoring: Exploitation of social media to gather citizens subjective response to noise

Subjective response to noise is probably the most important goal in environmental acoustics. Traditional surveys have the drawback of high cost deriving from its development and execution, the limited number of participants, and the duration of the surveying campaign. The main drawbacks of online surveys are the low participation, or the self-produced bias that concerns about the topic can raise. In both cases, the process of designing questionnaires, implementing the survey, and analysing the results can be long, expensive and ineffective to monitor changes in the short-term. With the creation of Online Social Networks (OSN), people have changed the manner they communicate and use technology. Nowadays, people can provide information regarding their likes, opinion and discomfort about any topic, including noise, in just a few minutes with their smartphone. These Internet opinions can be analysed automatically using machine learning and Natural Language Processing techniques to get insights that can help in the early detection of noise problems, or in the prior assessment of action plans. This information can be significant helpful in addressing noise management by local authorities and stakeholders. The purpose of this paper is to present a novel methodology, based on machine learning, allowing for the gathering and processing of OSN text data, enabling the generation of a system able to exploit the data to detect noise complaints and to classify them by source. This methodology has been piloted in a case study using Twitter, and the main results are presented.