The Stack Archive

Algorithm identifies artificially promoted Twitter memes and hashtags

Thu 9 Jun 2016

Twitter app

Researchers have created an algorithm which can pick up on commercial agenda in Twitter hashtags and memes by gauging whether the content has received artificial or organic attention.

The research team, based at the University of Southern California, explained that typically the social media site tags sponsored and paid-for tweets, but hashtags and memes are often neglected and promoted freely online and via TV.

‘If we can learn the characteristics of promoted campaigns, then maybe we can detect and even predict them,’ project lead Emilio Ferrara told New Scientist.

The algorithm has been trained to recognise what a promoted meme, labelled as such by Twitter, looks like and how it is distributed. Having learned this behaviour, the algorithm is then applied to unlabelled promotional campaigns.

The team found that the system was 95% accurate in its deductions, leaving out only one in 20 promoted memes.

Google software engineer Jacob Ratkiewicz told New Scientist that he was impressed with the technology, which was showcased at the Tenth International AAAI Conference on Web and Social Media, in Cologne last month. He noted how remarkable the algorithm was in its capacity to spot when human actors had been paid to retweet a promotion.

‘What this study is doing is looking at real human activity and trying to tease apart the organic human stuff from the promoted stuff,’ said Ratkiewicz

Christo Wilson, a researcher at Northeastern University, argued, however, that the software would still struggle with some types of social media advertising.

‘Let’s say that Shell paid Kim Kardashian to tweet something… That looks organic. She could have tweeted about her dog or Shell oil. The way it looks and the way it spreads look the same,’ he commented to New Scientist.

Wilson admitted that the algorithm would still be a great help with targeting ‘low-hanging fruit.’


advertising news research Twitter
Send us a correction about this article Send us a news tip