While not overtly using Twitter and Facebook messages to make predictions on the GOP campaign, a number of media outlets are investigating the use of sentiment analysis to gain insight on the public pulse. James Rainey's article in the Los Angeles Times describes attempts by Politico, which analyzed Facebook status updates, and the Washington post, which used a Twitter tracker, to draw conclusions about the various GOP candidates and how the social media public perceived their chances.

The analysis of social sentiment uses natural language techniques to explore emotions such as "happy" "angry" and "sad" in text-based messages. Generally, IT analysts may use sentiment to draw conclusions about the effectiveness of a marketing campaign or to rate enthusiasm for a particular product. Small and midsized businesses (SMBs) often gravitate towards analyzing sentiment because of the accessibility of free or inexpensive tools that filter and analyze tweets and status updates, giving them an entry point into social media analytics.

In spite of media interest in crunching sentiment as part of the political news, criticism has followed the practice. As discussed by Megan Garvey in a different Los Angeles Times article, interpreting language isn't that clear-cut. Examining work by Kanjoya, a San Francisco company with an algorithm that analyzes Twitter feeds for emotion, the Times reported that "joy" can be expressed at both a candidate's rise as well as his downfall. The intent of a particular statement seems to be key to interpreting emotion, and an emotion like irony is difficult for a machine to interpret.

For SMBs, agreeing on the interpretation of sentiment is only part of the solution. Additional metrics are needed--metrics that should assist in making more accurate assessments and spur action. This is where the free sentiment tools often fall, as their results are not easily integrated into more comprehensive business analytics tools.

What IT analysts at SMBs can learn from the media's foray into sentiment analysis on the campaign trail include the following:

• Social media sentiment may or may not represent the company's target demographics;
• Sentiment by itself is a singular metric and only useful when taken into context with metrics that spur action;
• The technology is new and imprecise, particularly when trying to interpret complex emotions such as irony; and
• People simply like to be a part of the social chatter, which may or may not be meaningful.

The last point is perhaps best exemplified in Rainey's article, which cited the Washington Post's @MentionMachine feature as finding that "momentum in S. Carolina" was associated with Romney 183,000 times on Twitter just prior to the South Carolina election, and wondering what meaning it might have. As later seen, that lone metric did not appear to be very meaningful at all.