Twitter Sentiment Analysis: Bridging Computer and Natural Language
By Rob Scott
November 12, 2012
November 12, 2012
- Tone in various forms of electronic communication has often led to confusion between friends, colleagues and family members — especially when the intended tone is sarcasm.
- “Now, as more people are sharing their opinions with casual acquaintances and strangers on social media sites like Twitter and Facebook — rather than in private text messages to people who know their senses of humor — the sarcasm disconnect is even greater,” notes the Wall Street Journal.
- Data miners are increasingly met with this challenge. “Sarcasm is proving to be an obstacle for the academics and marketers who create computer programs to analyze massive pools of online chatters to gauge public opinions about products and politicians,” explains the article.
- USC’s Annenberg Innovation Lab has launched a Twitter Sentiment Analysis project that “unites linguists, sociologists and computer scientists to try to build a modern-day lexicon for computers to read and interpret huge chunks of data provided by the millions of people who share their opinions online,” reports WSJ.
- The project has turned to politics to analyze data. For example, the lab’s computer has analyzed more than 40 million tweets involving candidates and hot-button issues. Additionally, more than 50,000 tweets have been manually analyzed to determine sentiment.
- “Many of the ETC member companies are using sentiment analysis from companies like Crimson Hexagon to guide their marketing and social media efforts,” explains ETCentric contributor Phil Lelyveld. “This USC research program will improve the accuracy of the analytics that they use.”
No Comments Yet
You can be the first to comment!
Leave a comment
You must be logged in to post a comment.