Opinion mining and sentiment analysis for policy making

Abstract

When implementing policy getting feedback is crucial for companies and governments. Many organisations face the problem of having large amounts of data as a result of marketing and research. Traditional techniques such as surveys and interviews demand large amounts of time, work and thought. Industries and Governments require solutions to the problem of having to deal with large datasets. Within this unstructured data, there is concealed information which could help the entity understand how exactly the customers or citizens are feeling towards any given subject. Such data can be better managed using the concept of data mining, which is the process of uncovering patterns and significant relationships in large sets of data. Sentiment Analysis is a subset of data mining technology that can be used to extract structured data. This structured data consists of interpretations of the emotions the writer is feeling. The concepts applied in this project will shed light on how applicable Sentiment Analysis techniques are in improving customer-company or citizen-government relationships. This Final Year Project includes the literature research on Sentiment Analysis techniques. Furthermore Sentiment Analysis techniques were further explained in context of how it has been used and for which purposes. Supervised learning techniques were applied to analyse opinions on a number of controversial topics currently being discussed. These opinions were inputted by participants on a web application that was developed. This eventually yielded rating results that gave indications to how positive or negative the text being analysed was. The primary results of the experiments carried out indicate that Opinion Mining and Sentiment Analysis techniques are in fact useful at providing an insight to the writer’s emotions and opinions. This could potentially help in improving business and governmental policy making processes.