Accurately analyse what people say on Twitter, and you can respond to customers, crowdsource design or predict the stock market. Seth Grimes from analytics consultancy Alta Plana gives his tips.
Know your goals
"Some people use sentiment analysis without any notion of the things they'll learn," says Grimes. "What are you going to do with that info, and have you collected it in a form that allows you to make a decision based on it?" Assign someone to collate data, rather than checking in now and then yourself.
Pick your tool
"You can use software toolkits, including open-source ones, to program if you have the skills," says Grimes. "Most people want something prepackaged." He recommends business analytics such as SAS for the high-end, Radian6 for the middle market, and Tweetfeel for those just starting out.
Run a test
Grimes suggests a way to assess the strength and nuance of your tool: "Enter the name of someone who's died and see if they rate as negative a tweet that says, 'I'm so sad that this person died.' That's actually a positive sentiment. It's a good way to turn up weaknesses of some of these tools."
Add a human
"You need to combine the strength of a computer with the capabilities of a human," says Grimes. Add in social customer-relations management or respond to the tweets as well as monitor them. "Winnow down to the people who are actually customers," he suggests.
Thank You : wired.co.uk
Know your goals
"Some people use sentiment analysis without any notion of the things they'll learn," says Grimes. "What are you going to do with that info, and have you collected it in a form that allows you to make a decision based on it?" Assign someone to collate data, rather than checking in now and then yourself.
Pick your tool
"You can use software toolkits, including open-source ones, to program if you have the skills," says Grimes. "Most people want something prepackaged." He recommends business analytics such as SAS for the high-end, Radian6 for the middle market, and Tweetfeel for those just starting out.
Run a test
Grimes suggests a way to assess the strength and nuance of your tool: "Enter the name of someone who's died and see if they rate as negative a tweet that says, 'I'm so sad that this person died.' That's actually a positive sentiment. It's a good way to turn up weaknesses of some of these tools."
Add a human
"You need to combine the strength of a computer with the capabilities of a human," says Grimes. Add in social customer-relations management or respond to the tweets as well as monitor them. "Winnow down to the people who are actually customers," he suggests.
Thank You : wired.co.uk
Posts
0 comments
Post a Comment