Emotion driven business: how sentiment analysis can be applied to social media content to extract business value

Expressing emotions is one of the key innate features of humanity. The capacity of articulating feelings exists since birth and grows in complexity as the individual becomes more self-aware. The rise of social network sites (SNS) has however amplified this phenomenon, modifying current communication methods and introducing new ones. The thesis consists of the description of existing sentiment analysis algorithms of Facebook posts and their potential business applications. In the Facebook environment, users interact most with others by liking posts and commenting them. Studies have demonstrated that on SNS, consistently with social sharing theory, hearing about a friend’s troubles causes friends to reply more with longer and more emotional and supportive comments (but also containing more negative feelings, presumably phrases like it sucks or I feel you), while giving out less likes. Posts with positive feelings, on the other hand, receive more likes, and their comments have more positive language. Emotions play a huge role in business and in the success of enterprises. They influence teamwork, customer satisfaction, manager-employee relationships, and employee retention. Plus, the brain’s emotional state affects decision making, planning and negotiating, and creative thinking. Consequently, understanding both feelings of customers and employees expressed over social media could give companies a competitive advantage over competitors. Text communication via Web-based networking media, however, can get somewhat overwhelming to analyse. Every second, a massive amount of unstructured data is generated on the Internet due to social media platforms. This problem can be solved with sentiment analysis. Sentiment analysis, one of the applications of AI, is the activity of using statistical methods and learning algorithms to analyse any written text and extract subjectivity and polarity – respectively, whether the text contains any non-factual information and the positivity of its tone. Two approaches to sentiment analysis are presented: the first one based on Microsoft Azure Cognitive Services for Language, the second one based on the VADER learning model. Both approaches are based on Python programming language. Results of the models show a negative correlation between sentiment and number of shares and likes for the first one and an accuracy of 0.44 for the second one. While describing potential applications of any algorithmic results, ethical and sustainability problems need not to be undervalued as innovation can be truly sustainable and lead to the renewal of the national socio-economic system only if it considers the synergy between demand of citizens and environmental and ethical risks. Examples of applications of sentiment analysis in a business context can include harmful ones – like in the case of Cambridge Analytica – or positive ones – such as use of sentiment analysis for customer retention, understand investment tendencies, offer better online psychological support and understand feelings of employees. In all examples, sentiment analysis provides immediate and accurate results compared to traditional survey methods and gives to companies a cutting-edge competitive advantage.