02 October 2013
Twitter unlocked by Topsy
Topsy is a new tweet analysis service that can be an extremely useful tool for reputation managers. But relying totally on its sentiment algorithmns is unwise ... false negatives abound and as we all know, Americans and algorithmns don't do irony.
One of the features of Twitter is its transitory content. Tweets emerge and disappear like rain drops falling on a pond.
From a reputation management point of view, this is both good and bad. On the good side, any bad chat soon gets buried by tomorrow's hot topic. On the not-so-good side, it has been near impossible to map tweeted opinions and moods over time, because Twitter's archive, search and analysis functions are weak.
Topsy (www.topsy.com) has changed all that. On its website you can trawl through all the supposed 540 billion tweets uttered since the first one sounded in 2006, to see how your brand or product has fared over time. In fact it is so good that Topsy has been embraced by Twitter itself.
If you want to know more, like how your brand fares among different categories of user, or in different regions, or how recipients respond to the way a message is phrased, Topsy can provide this service to you for a fee.
For those of us who don't have the budget of Coca Cola or a US presidential candidate, Topsy provides a search and tone analysis function free of charge, for tweets made in the previous month.
Check your brands or company name out on it. Or sign in for the full monty, free for the first 14 days. Results can surprise.
As a matter of interest, we checked out 100% Pure New Zealand, a slogan that's been controversial locally.
Topsy 12-month sentiment score for 100% Pure New Zealand (screengrab)
It's clear from Topsy's analysis why NZTE is sticking with it for promoting New Zealand as a tourist destination. The tweets are overwhelmingly positive and most of the negatives appear to be misinterpretations of sentiment by the Topsy algorithmn.
Some examples of tweets that Topsy appears to classify as 'negative':
@mayaCR05 [Maya 492 followers]: Mount Taranaki => The Lonely Mountain? #TheHobbit #NewZealand@PureNewZealand@theoneringnet http://t.co/uqz1uyK0
@emirarnas Emir [444 followers]: @MayaCR05 @PureNewZealand @theoneringnet Far over the misty mountains cold
@blackwolf409 [Blackwolf409 103 followers]
@EmirArnas @MayaCR05 @PureNewZealand @theoneringnet To dungeons deep and caverns old
@purenewzealand [Tourism New Zealand 48K followers]
It's time! The Hobbit: The Desolation of Smaug trailer is here. http://www.youtube.com/embed/mbOEknbi4gQ @TheHobbitMovie
In other words, don't take Topsy's sentiment analysis too literally. In the examples above, words like 'lonely', 'cold', 'old' and 'desolation' appear to be classified as negative, when in context they are positives.
Then there are tweets that are genuinely negative, but which have nothing to do with 100% Pure New Zealand, like this stray bird from the Maldives.
@maldivesfg [Mv Future Gen | KKBK 831 followers]
Most corrupted minister in the Maldives Coup Government is @Ahmed_Adeeb! He trying to save his ass before #MvElection! @PureNewZealand
False positives are also likely, though we didn't find any in this exercise. Algorithmns (especially American ones) are always going to struggle with irony. For example a tweet saying: 100% Pure New Zealand dairy farms. Yeah right! Would doubtless be interpreted as a positive.
- Trevor Walton