Extracting Strong Sentiment Trends from Twitter

Twitter is a popular real-time microblogging service that allows its users to share short pieces of information known as “tweets” (limited to 140 characters). Users write tweets to express their opinions about various topics pertaining to their daily lives. With a total 175 million users and 95 million tweets published per day (as of September 2010), Twitter serves as an ideal platform for the analysis and extraction of general public sentiment regarding specific issues. The measurement of presidential performance is one domain where the analysis and extraction of general public sentiment is a large component. Currently, presidential approval polls are hand-measured by random telephone sampling of a small population. This technique is both timeconsuming and costly. Therefore, an automated way of measuring these polls from easily accessible public data would be immensely useful in reducing the required time and costs. This project explores an approach to automatically extract large-scale trends in the perception of presidential performance among the general public by analyzing tweets published on Twitter. Specifically, macro trends in strong approval and strong disapproval of presidential performance are extracted from tweets using a simple lexiconbased approach. The extracted sentiments are compared against a hand-measured presidential performance poll to measure correlation and determine whether strong political sentiments regarding presidential performance can be extracted from Twitter. From a natural language processing perspective, this problem is interesting because ...