Twitter has dominated headlines in the mainstream press since the turn of the year having first of all been tipped for an $11bn flotation in the US next year and more recently it has been reported that it posted a modest £16,500 profit in the UK in its debut filings.
First of all, I’d point out that the modest UK filings aren’t necessarily a fully accurate reflection of the true commercial success of Twitter here given that start-up costs don’t come cheap when setting up in a new market.
What I find more interesting, is a more subtle post on the official Twitter engineering blog from Alpa Jain, a Twitter software engineer, and Edwin Chen, one of its data scientists, which offered an insight into how the social network aims to improve its search capabilities through “real-time human computation.”
In the post they explain how Twitter is the destination for its audience to discover what’s happening now and that this creates difficulties in returning results, from both a search and advertising perspective, as “search spikes” are oftentimes short-lived.
This makes it difficult for the “interest network”, as its employees often refer to it as, to properly contextualise its results plus Twitter’s offering means that there’s only a small window of opportunity for its search engine to learn the context of these queries.
Jain and Chen quite rightly point out in the post: “How would you know that #bindersfullofwomen refers to politics, and not office accessories, or that people searching for “horses and bayonets” are interested in the Presidential debates?”
As a response, Twitter is now ramping up its “human evaluation” to improve its real-time search results and serve better ads against them by pairing its software with human understanding.
The post explains how it works: “First, we monitor for which search queries are currently popular… As soon as we discover a new popular search query, we send it to our human evaluators, who are asked a variety of questions about the query.
“For example: as soon as we notice “Big Bird” spiking, we may ask judges on Mechanical Turk to categorize [sic] the query, or provide other information (e.g., whether there are likely to be interesting pictures of the query, or whether the query is about a person or an event) that helps us serve relevant Tweets and ads.”
As with any search engine, Twitter then indexes the context of the search term and indexes it for posterity – this then improves the accuracy and relevancy of its future results.
If this proves effective, and Twitter continues to build on its 200 million-strong audience, I’d argue that it could prove an attractive proposition to marketers looking to find alternatives to Google, Bing and Yahoo when deciding on their search strategy.