For the fulfillment of the first general objective of this doctoral thesis, a study was carried out in the city of Malaga (Spain) to quantify the effect that the communication of noise pollution information has on the perception of annoyance associated with noise. The experiment focused on noise emissions produced by recreational activities since they are a problem of growing importance for urban managers in Southern European cities and one of the most important causes of disturbance for city residents. The research was done using an online survey-based methodology. First, a questionnaire was carried out to find out the population's noise annoyance before announcing the existence of a leisure noise monitoring campaign. Then, the project was disseminated and a platform for displaying acoustic information in real-time was made public. At the same time, another questionnaire was conducted to check whether there had been any changes in the annoyance that could be associated with the communication and dissemination of noise pollution information.

In terms of the methodology designed, it has been demonstrated that PRE-POST evaluation using online questionnaires is a valid system for measuring the variation in the annoyance generated by leisure activities before and after carrying out a series of actions that could modify the perception of a noisy activity.

This methodology has shown that the online platform for displaying noise data has not affected the annoyance associated with leisure noise. However, the results point in the direction that being aware of the existence of the monitoring network installed in Malaga increased the annoyance of citizens by 1.3 points. This increase in discomfort could be related to an increase in awareness of the noise problem, since by deploying a monitoring network the citizens could have perceived that noise levels were high enough to have to be controlled. On the other hand, given that the survey was freely accessible, the most affected people could have accessed the questionnaire more frequently, causing a general increase in annoyance that does not represent the overall population.

It was observed that the city's anti-noise associations spread the survey through their own communication channels, resulting in changes in the distribution of the sample between the initial survey (by invitation) and the final one (free access). This highlights the importance of exploring the public media in which online questionnaires are shared in order to control possible biases associated with the distribution of the sample.

In addition, it has been demonstrated that the population which belongs to these associations presents different patterns of annoyance from those of standard citizens, with a feeling of discomfort 1.5 points higher towards leisure noise. This highlights the need to include a question asking about this factor.

From the point of view of the survey dissemination and its relationship with participation, the objective of getting wide coverage of the project in the media has been achieved, although this did not translate into high participation rates on the surveys.

Although leisure noise is an issue that both citizens and environmental managers recognized as one of the most disturbing sources in Malaga, the participation obtained did not meet the expectations. Despite this, the use of online surveys has proved to be a valid tool to increase participation, since the number of citizens who completed the questionnaire compared to other noise surveys previously conducted in the city through face-to-face interviews was increased by 40%.

On the other hand, a completion rate of 80% was obtained in the questionnaires, well above the 50% reported in other investigations with online surveys, which verifies that the length of the survey developed was adequate.

From the experimental design point of view, it has been observed that taking advantage of the media coverage during the project launch to announce the existence of the survey has not been the most optimal method for the experiment. Many people accessed the survey without having a reasonable time to explore the application of acoustic data visualization, so they had to be excluded from some of the analyses because they did not meet the requirements of the experiment. In future research, this must be taken into account to attain greater public participation, which is in itself already difficult to obtain.

A methodology has been proposed to obtain, classify, and analyze texts for the purposes of investigating noisy activities coming from online social networks. By means of machine learning and natural language processing techniques, a detector and classifier of noise complaints written on the social media network Twitter was implemented. In addition, a method for the analysis of the complaints detected was proposed. This method allowed the detection of noisy events, to know the origin of the disturbing sound and predicting them some time in advance.

The investigation has confirmed that some in the public publish information about the noise sources that negatively affect them in alternative channels to official ones, such as online social networks, to express their discomfort. This information can be acquired through a search based on keywords and the APIs of some social networks, such as Twitter. This makes it possible to obtain a large amount of data for use in evaluating the level of annoyance social media users experience due to noisy activities.

Concerning data preparation, it has been demonstrated that the implemented text preprocessing method, in which slang and diminutive words are normalized and grammatical mistakes are corrected, is valid to increase the dimensionality of short texts. This has allowed a greater lexical richness in the content of the texts, facilitating the extraction of representative text features and thus improving the performance of the implemented classifiers.

Since it is the first time NLP techniques are being used in the detection of complaints about noise, a set of classes has been defined to use the texts coming from social networks in supervised machine learning algorithms. The process carried out to validate the defined classes, based on measuring the degree of agreement between different annotators, has allowed demonstrating that the created categories were correctly specified and that they can be used optimally in this type of classifier.

The combined use of textual features computed using NLP techniques, such as n-grams, morphosyntactic features, sentiment features, and embeddings, have been valid for training an automatic text classifier capable of detecting noise complaints, obtaining a performance similar to the shown in the state-of-the-art investigations. In addition, the expansion of the taxonomy developed in other investigations using semantic web technologies and lexical-hierarchical dictionaries has been proved to be an ideal method for the development of a system capable of classifying noise complaints by sound source according to the words present in them.

For the performance tests of the system, a methodology that implements stratified cross-validation techniques and weighting coefficients to each of the existing classes in the data has been proposed. This approach has managed the imbalance between the classes in the training data of the system. In addition, the fact that 85\% of noise complaints have been correctly detected by the classifier and 86\% of the complaints were adequately assigned to their sound categories demonstrates that the methodology implemented for the analysis of these texts is valid for obtaining information on the negative perception of noisy sources present in social networks.

Additionally, the use of the classifiers designed together with ARIMA intervention techniques makes it possible to detect time periods in which anomalous noise-related annoyance situations occur, as well as to know the general origin of these complaints.

The approach used sought to know the specific origin of the noisy event detected, based on the semantic analysis of complaints to know the frequency of use of words in the texts, has been tested as a valid method to know the specific sounds to which noise complaints detected during a noisy event refer to.

In addition, the use of statistical process control and time series analysis techniques has enabled the creation of an alarm system capable of detecting specific annoying sound events by counting the number of complaints and the presence of specific words.

The systems developed to fulfill the second objective of the doctoral thesis demonstrate that, indeed, information about noisy activities can be obtained from social networks. In addition, these tools are of interest for urban managers to detect problems related to noise in real time through the information that citizens share through online social networks. On the one hand, the alarm system could be trained to detect events related to recreational noise and detect non-compliance in the closing hours of nightclubs and pubs. In addition, the system for detecting sound events and the specific source of complaints could be used for rapid action in the face of acoustic situations that disturb citizens, and to measure the effectiveness of these actions based on the decrease in complaints published after they have been carried out.

Technological development and the advancement of management policies have made it possible today to obtain a large number of different types of data in a public manner. Citizens, in addition to sharing texts about their attitudes toward noise in online social networks, share multimedia data that can provide information about the acoustic environment in a city. Furthermore, the growing importance of research that studies the effects of noise on human health has led to an experiment carried out in London to analyze the influence of different sound sources on public health.

The experiment has shown that in a large city, there are enough public data sources to carry out research that makes it possible to know if noise or the presence of specific sounds impacts the health of the urban population.

The methods implemented for estimating the percentage of the population exposed to different noise levels and evaluating the presence of sound sources in specific areas of a city have been proved valid to study the impacts of noise on the population's hypertension.

In addition, the results of the analysis done in this study have allowed to reach some conclusions:

First, it should be noted that exposure to railway noise did not seem to have an effect on hypertension in any of the models calculated, despite the existence of previous studies showing the opposite. This could be because the level of spatial aggregation masked small areas where this sound source had a high impact, or because areas near railways were protected by acoustic screens that reduced the impact of noise on the resident population. Similarly, night-time noise levels have not been relevant, presumably because the percentage of the population exposed to noise in this time span was not high in each MSOA.

Second, the results of the experiment suggest that socioeconomic factors are the main source of information when studying diseases at the MSOA level. Furthermore, although it is known that economic power affects the prevalence of hypertension, the study sheds light on the limits of traditional noise exposure models, which consider only means of transport to measure the effects of noise pollution on the well-being of citizens. In fact, the inclusion of variables that take into account a greater diversity of sound sources, both positive and negative, allows for a greater understanding of this disease when there is no possibility of knowing the exposure to noise in an area.

The results could lead to new methods of noise management and could be of interest to urban planning professionals. Since the results suggest that exposure to certain specific sound sources improves hypertension, the results open the door to new studies to measure in detail the impacts of these sounds on health in order to design acoustic environments in cities that are beneficial to the urban population. Furthermore, the usability and efficiency of social media platforms to measure the presence of sound sources has been demonstrated, which could complement the information given by the sensor networks with the capacity to detect sound sources in this type of study.

On the other hand, it has been shown that the importance of social media variables is boosted when studying the effect of noise on hypertension in some neighborhoods where traditional metrics, such as the percentage of people exposed to traffic noise, are unable to explain the disease due to a lower flow of vehicles or specific conditions of these areas, something that in the case of the experiment has occurred in the most wealthy neighborhoods of London. This is again of interest to urban planners, who could thoroughly study the sound environments in these upper-class neighborhoods and replicate them in other areas with the intention of improving the quality of life and health of the residents.