Abstract:
Data Analytics is the science of extracting meaningful information from large amounts of seemingly meaningless data. This information in turn helps to monitor trends, predict outcomes and model business decisions based on facts. Our work is based on the subfield of Sentiment Analysis and Location Estimation. Using machine learning algorithms and natural language processing, we have created a tool that allows businesses to perform sentiment analysis and location estimation on social network data.
Our software fetches live data from Twitter and passes it through several stages of processing to extract sentiment and other accompanying attributes related to the data that are relevant to businesses. After the cleaning and spam removal phase in which 85% of all spam is removed, our sentiment classification algorithm not only assigns sentiment to the text as a whole (as done by other solutions), it is also able to extract topics within a post and the specific sentiments associated with those topics, so that we can see the main features of the entity which are responsible for the assigned sentiment. Sentiment classification is achieved by our tool with an accuracy of 72%. Moreover, as various consumer demographics are also necessary for a complete analysis we provide a unique algorithm that estimates the location of the user within 80 miles in up to 55% of the cases, whereas geo-tags are only available in less than 1% of tweets. This, coupled with our topic based sentiment analysis and other statistics, are presented using relevant and informative visualizations in our tool, providing a complete solution for anyone seeking to extract opinion from social media.
