Tuesday, February 26, 2013


NetSentiment is an opinion mining service that analyzes data in real time from news and social media, applying human language technologies.

It turned out to be more accurate in predicting the outcome of the Italian elections that traditional tools based on opinion polls.

There were a few things that the opinion polls missed as well as the press and the politician who read the press and the polls.
In the last year during the Monti government and while Berlusconi had retreated from active politics, the traditional media (press and TV) had dispensed hunanimous discredit on him.
This influenced the opinion polls, since apparently many people were afraid to admit they were voting for Berlusconi, but in the social media, they would voice more freely their disappointment for the left coalition and react strongly.
The disappointment for Bersani shows up in the low score, well below Grillo and Berlusconi. Despite this, still a large number of people voted for the left coalition just for a sense of duty. This explains why the votes for the Bersani coalition are higher than the Bersani NetSentiment index.
Grillo has an immense popularity on the net and it was surprising that the NetSentiment index was so low until ten days ago when it jumped to first place, possibly in connection with his Tsunami Tour which filled the Italian squares.
The NetSentiment data on the day before the elections (now they are different), show a quite accurate picture: Grillo 25%, Berlusconi 22%, Bersani 19%.

The first two are very close to the actual percent of votes of their parties. The election results for the PD party (25%) are sligtly higher than those of Bersani, for the reason mentioned above.
This looks similar to what happened in the Catalan elections for the PPC candidate.
So in future election analysis one should try to consider the effect of party popularity on each candidate.
For both Berlusconi and Grillo, candidate and party coincide but not for Bersani.

No comments: