Development of a monitoring dashboard for sentiment and emotion in geolocated social media

Jorge García-Castaño. (2017). Development of a monitoring dashboard for sentiment and emotion in geolocated social media. Final Career Project (TFG). ETSI Telecomunicación, Universidad Politécnica de Madrid.

Abstract:
XXI century has brought to the world new communication standards, such as voice calls, internet connection, and a plethora of multimedia content. Once these elements became mainstream, they also became a main economic target for several organisations. This is why, nowadays, social networks are enterprise's activity center towards clients. Although Twitter has an implicit limitation (140 characters) compared to its competitors, there is an enormous amount of information that can be extracted for economic use. How people feel about a certain event or social movement are good examples of it. Thus, a simple and summarised representation of general emotions and sentiments is becoming a necessity in the short term. One of the most striking technologies in this issue is the sentiment analysis. Although it is still a growing and developing field, it is already a crucial part of businesses around the world. However, to make sense of sentiment, it cannot be analyzed in isolation. As with other aspects of social analysis, it needs to be visualized in context. Many tools enable temporal and personal information. In other words, one could ask questions such as "How are users feeling?", "how is this specific user feeling?" and "who's the most positive and influential person??. However, one aspect that is often ignored is location, so it is not possible to explore emotions in certain areas. e.g. the emotions of students in Ciudad Universitaria. Hence, the main focus of this thesis will be to develop a dashboard to monitor public sentiment emotions in geographic We will leverage two open source technologies developed at GSI: Senpy, a framework to develop semantic sentiment and emotion analysis services; and Sefarad, a framework to create dashboards to explore date from semantic ElasticSearch and SPARQL endpoints. The development of this thesis will consist of several sub-tasks. First, it will require the development of a heat map widget for Sefarad to display sentiment and emotion analysis in maps. We have named this widget HappyMap. Once the HappyMap is integrated in Sefarad, we will implement a dashboard that combines it with several existing Sefarad widgets. e.g. the Tweet List widget to show the list of tweets in the area, and the Cheroff Faces plugin to display an aggregate of the global emotion. Additionally, we will design and implement a pipeline to extract tweets, annotate them with sentiment/emotion and load them in the semantic backends. This whole project will be deployed by means of docker containers, which eases the integration of different services (e.g. ElasticSearch) and its deployment. In summary, this thesis will use the following technologies, libraries and languages: Python, ElasticSearch, HTML+Javascript, Leaflet, Polymer, Senpy, Sefarad, Docker and Luigi.