Sentiment Analysis and its Application in Managerial Decision Making

Abstract

Many cities around the world are aspiring to become Smart by deploying various information and communication technologies in order to support decision making. Smart City initiatives are often an interplay of local authorities and businesses involved in this industry.

The opinions of ordinary citizens are often given little weight. This could be due to the costs and difficulties associated with collecting and analyzing these opinions. Many cities use surveys, but this data collection method has many limitations, such as the enforcement of a closed set of questions and answers or limited temporal scope.

Free-form text published on social media could be used as a complementary source of citizen opinions. However, processing these data is even more difficult than evaluating survey results. Therefore, this dissertation proposes a novel social media analysis framework for municipal decision making. The framework consists of three main components. First, it uses topic modeling methods to identify topics discussed by social media users at a given location. Second, it uses sentiment analysis to determine the degree and orientation of the sentiment polarity of social media content. Third, it aggregates topic and sentiment information to provide an overall high-level view on the challenges and opportunities the municipality is facing.

The aggregation component is the main methodological contribution of the dissertation. It uses fuzzy sets to capture the uncertainty stemming from different people having different opinions on the same topic. The framework also determines the level of positive and negative opinion expressed towards each topic as a degree of similarity between the fuzzy set representing sentiment towards a specific topic and fuzzy sets representing positive and negative opinion.

The functionality of the framework is demonstrated on synthetic data and more importantly on real-world data extracted from Twitter for two Czech cities: Ostrava and Brno. Several conclusions useful for municipal decision making are drawn from the analysis. The framework is then compared with a naive approach, and it is shown that it provides more information. Finally, a the dissertation presents a simple web application that illustrates how the framework results can be presented in a user-friendly form.