Twitter took the longest time to monetize it service using an advertising model. Back when Facebook was making a lot of money by doing ads, Twitter chose to focus on growth and product. Last year, Twitter’s CEO Dick Costolo spoke about how their ad program was doing well, especially on mobile. One of the product features that the Twitter’s team leverages is its real-time search of trends and keywords.
In a post on its Engineering blog, Twitter’s engineers talk about how the real-time aspect of Twitter’s search is too complex for algorithms to make sense of. Twitter has built algorithms to detect spikes in traffic centred around keywords and hashtags, but sometimes these keywords and these hashtags cannot match the context, like human beings can. Hence, the Twitter team uses humans to sort out the problem.
So this is how it works. When there is a search spike in terms of keywords and hashtags, Twitter uses a system called Storm which tracks all this data in real time. The next bit is interesting. Twitter then uses Amazon’s Mechanical Turk system to sort this data and attach context (relevance to certain situations). Mechanical Turk is an crowdsourced platform which employs humans to attach all the context. The humans whom Twitter addresses as ‘Judges’ also associate spiking search terms with images, categories and so on, thus giving meaning to the data and allowing Twitter to serve the perfect advertising to people who are searching for the term. Hit the story source link below for the complete dope on the subject.
This is brilliant as it gives Twitter revenue in advertising and at the same time makes their real-time data search very useful for people. I think, Indian startups in a similar space should also check out Mechanical Turk if they have some complex problems which cannot be solved by algorithms.
Via | The Verge
Story Source | Twitter Engineering Blog
Image Source | H+ Magazine