AbstractToday, social networks and e-commerce platforms occupy a huge place in our society. These media are an important source of messages from users expressing an opinion or sentiment whether about an event or a commercial product. These subjective messages contain a wealth of information that is difficult to analyze manually and, for several years, a discipline has emerged that seeks to automate the analysis of this data: sentiment analysis. The Twitter micro-blogging platform, by its number of users and its number of daily messages is an interesting resource to work on this kind of content. In this document, existing sentiment analysis techniques are presented and various
publications in the field are detailed. The main role of this document is to
investigate the ability of an approach coupling neural networks and n-grams of messages posted on Twitter to provide good results as part of a sentence level sentiment classification. To achieve this objective, a pipeline was set up to cover all the operations required to carry out this experiment: data collection and cleaning, dataset preparation and training of the neural network.
|Date of Award||19 Jun 2018|
|Supervisor||Wim VANHOOF (Supervisor)|