- Understanding aviation noise
- Improving impact management
- Understanding spatial variations
Social media is becoming an integral part of people’s lives. They use these online social spaces to exchange information with their friends, collect data on various subjects and also to discuss, debate and complain about different aspects of life. This makes social media a rich source of information, and this includes information related to the messages exchanged (textual data), and information related to the connections people have with each other. Social media is also used as a place for organising political and social activities like online campaigns, demonstrations and to express annoyance and disturbance.
On the other hand, can we understand if people feel annoyed or uncomfortable in their own environments due to external reasons like airplanes flying or noise generated by airport activities? Can we sense what this means to their quality of life? More importantly, can we do that without invading their personal lives, without following them around and without asking for more information other than what they have already shared by themselves?
The current work in ANIMA tries to bridge the gap and provide innovative tools based on Artificial Intelligence (AI) and Machine Learning (ML) to find the relevant discussions on social media that the users have already posted. We can actually follow discussions of previous years or ones that take place at this moment. Then we analyze the posts to find their polarity or sentiment (semantic analysis) by training learning models to recognize polarity in statements. Based on that, we try to estimate if the users are having negative of positive statements towards the annoyance generated by the airport (when and if it is the case) and we chart the temporal and spatial evolution of this.
The prototype analytical system that we built, gives us very encouraging results so far, with an accuracy exceeding 75% and while working without the need for previous knowledge (i.e., no pre-labelled data are required).
The prototype will be published as an open-source toolkit available to anyone that would be able to:
- use social media and other user-generated content on websites to chart the sentiment of people about the possible annoyance caused by the airport activities in their lives;
- extract an adequate number of posts, that allow us to work with some certainty for our results, although further analysis is needed to explore bias in the data and to be able to generalize the conclusions;
- draw results based on the comparison internally of the data and we do not try at the moment to correlate with external sources;
- analyse posts both in real-time and offline and provide insights towards the users’ feelings; this toolkit is not tied to a specific airport or even a particular language. We can use it to collect and analyse information and generate visualisations of this information.