By Jonathan Fildes
Technology reporter, BBC News
The researchers studied films such as Avatar
Micro-blogging service Twitter can be used to predict the future box-office takings of blockbuster films, according to researchers at Hewlett Packard (HP).
The computer scientists studied 3 million messages - known as tweets - about 25 movies, including Avatar.
They found the rate at which messages were produced could be used to accurately predict the box office takings before the film opened.
Further analysis of the content of the messages could predict ongoing success.
"Our predictions were incredibly close," Bernardo Huberman, head of the social computing lab at HP, told BBC News.
For example, he said, the system predicted that zombie film The Crazies would take $16.8m in its first weekend in the US. It actually took $16.06m.
The team forecast that romantic drama Dear John would take $30.71m in its first US weekend. It took $30.46m.
The unpublished research has been posted on the Arxiv website.
The team were able to make their first-weekend revenue predictions by analysing the torrent of tweets about a particular film in the run up to its release.
"We developed algorithms to analyse these tweets and measure the rate at which they were produced, said Dr Huberman.
"Our intuition was that the faster people tweet, the more likely they are to go and see it."
The teams were then able to forecast the ongoing success of a film, including its second weekend revenues, by doing what is known as sentiment analysis.
This analyses the content of tweets and decides whether it is positive, negative or neutral.
"It's tapping into collective intelligence," said Dr Huberman, who carried out the work with Sitaram Asur, also of HP.
The team trained their system using Amazon's Mechanical Turk, an online tool that pays people to perform small tasks that computers would struggle to complete.
"We got people to classify tweets and we used that to calibrate the sentiment analysis," explained Dr Huberman.
Again, the system tracked the fortunes of movies and outperformed other predictive systems such as the Hollywood Stock Exchange, he said.
For example, analysis showed a boom in positive sentiment about the Oscar-nominated The Blind Side after it was released, but showed the opposite effect for New Moon, which initially sold well but rapidly lost viewers.
"Word of mouth builds audience," said Jan Saxton, vice president and senior films analyst at Adams Media Research.
She highlighted the film My Big fat Greek Wedding, which she said became the "film of the year" in 2002 by recommendations.
"If word of mouth becomes a faster, more effective marketing tool, then the effect on the movie business could be profound."
Both Dr Huberman and Ms Saxton said that the demographic of Twitter - which tends to be young, tech-savvy and reasonably affluent - may limit the utility of the system for analysis of some trends, such as those aimed at children.
However, Dr Huberman believes it could be of use in forecasting other trends, such as how well a gadget or product will sell.
Elements such as sentiment analysis are also being used by other groups.
For example, it is being used by an organisation called Tweetminster to monitor the UK general election to work out whether online buzz correlates with the winners.