Sentiment analysis is nothing new but is taking on urgency with the rise of mobile computing and the event enabled enterprise.
What is sentiment analysis?
Sentiment analysis is all about figuring out the attitude of someone speaking or writing. With so much being expressed through various social media across the Web, knowing the actual attitude behind the words is more important than what’s being transmitted.
We do this with software applications that allow us to use automation to track sentiments about products, brands and individuals and to understand whether they’re viewed positively or negatively. We analyze blogs, reviews, tweets and comments as broadly as they’re available.
How sentiment analysis works
The key to sentiment analysis is transformation of text into formats that applications can read. Not surprisingly, it involves using math to make decisions about the positive or negative aspects of text. But it isn’t easy to translate human feelings into data. Beyond straight forward speech, there is irony, sarcasm and slang, each presenting a tough challenge and many shades of emotional gray. It is the intersection of psychology and artificial intelligence.
Because of these challenges, getting this to work is tricky. Computer programs need to understand subjectivity and to be able to parse phrases to determine polarity (negative and positive meaning). Those same programs need to be able to figure out the intensity of subjectivity (eg. really strongly like or just like). Applications can try to figure out everything or just stick to the more obvious and call it good enough.
These sentiments are key to the new bread of ‘social mission control‘ centers being set up by major corporations. Not only can products and people be tracked, but individual aspects of an idea can be tested, as the NY Times described:
Using Newssift, a search for Wal-Mart reveals that recent sentiment about the company is running positive by a ratio of slightly better than two to one. When that search is refined with the suggested term “Labor Force and Unions,” however, the ratio of positive to negative sentiments drops closer to one to one.
We can not only perform sentiment analysis, but sub-segment analysis as well. This gives organizations the ability to test ideas in the marketplace without testing the entire brand. It allows for mid-course corrections on messaging or product and services before the full market impact is known.
Mobile technology stands to change the nature of sentiment analysis through the ability to request and receive feedback from exactly where and when a thought is occurring. Like with other social concepts, mobile significantly ups the ante for volume of information, location and time. Mobile gives an organization the chance to intervene while the customer is still unhappy and even still in the store or website.
Making it useful
- Sentiment alone rarely causes behavior. Behavior occurs when sentiment and other environmental factors combine.
- Knowing sentiment is good, but it only matters when action can be taken on what’s known. A result of sentiment analysis or an inflection in response should be considered an event that can be stand-alone or correlated with other events that have been identified by the organization as important enough to watch for.
- Watching for events means technology. An event processing application listens to channels that include sentiment analysis to decide when it matters most to act.
- Actions are controlled by rules that decide how correlated events will be treated by the organization. Responses can vary from a simple alert to the kicking off of workflows. Actions are where the rubber hits the road.
Without a system to manage measurement and action, all of the sentiment analysis in the world won’t make a difference. Like most things in our globalized, connected world, it isn’t what you know, but how you act on what you know.
Can you measure how your customers feel about you and your products and services? Can you respond when you find out?