Read more: Engines of the future: What鈥檚 next in internet search?
The future is being prepared now, so we may be able to tell where it鈥檚 headed if we can learn to read the web right
Web search holds the key to the future. So say the Time Monks, as the group that runs the project styles itself. This uses web crawler software to find 300,000 keywords in blogs, forums and chat rooms, then applies a to the text around each keyword, combining the results to predict future events.
Advertisement
So far the results of this 鈥渨isdom of crowds鈥� approach have been mixed. Yet in May Google saw fit to invest in , a Boston-based start-up that specialises in novel ways to relate the past, present and future. Its software collates web-based information about people, places and events, as well as the tone of news reports and posts from tweets, blogs and social media sites. Specialised algorithms then label the information, look for connections and attempt to plot the 鈥渙nline momentum鈥� for each event. According to Recorded Future, this can help predict events such as stock market trends, the launch of new pharmaceuticals, even terrorist attacks. Yahoo鈥檚 research lab in Barcelona, Spain, has developed a similar system called Time Explorer.
Yet search queries themselves could prove more valuable. With access to billions of these queries, search engine analysts have an unparalleled insight into the collective mind of the online community. In the last five years they have found evidence that figures for , and Israel, book rankings on Amazon, and the incidence of certain cancers are all mirrored by changes in the volume of web searches for associated terms. Studies by Google and with data from Yahoo have concluded that the volume of flu-related search terms mirrors the number of new flu cases in the US. By exploiting this search query data, service has proved able to pick up changes in the incidence of flu they are reported officially.
It may not be predicting the distant future, but it is certainly helping us pin down the present, says Nello Cristianini, a computer scientist from the University of Bristol, UK. This 鈥渘owcasting鈥� can be extremely useful, he says. For example, government agencies can use it to infer the state of society or the economy in near real-time, and the information is cheap and relatively easy to gather.
Cristianini believes this kind of data may be able to help predict an impending event such as a humanitarian crisis, and hopes to improve the technique鈥檚 reliability. He is testing an automated search that hunts for in Twitter posts in the UK. His suggest that interrogating Twitter can provide a useful complement to analysing search statistics.
Web search trends could also prove useful to business analysts, according to a in New York. It examined whether search data could predict box office takings at cinemas, video game sales and the chart positions of songs weeks or months in advance. The results suggest that in a few cases, search data is better than conventional indicators and so could be important in financial analysis, where even small improvements in the accuracy of predictions offer significant rewards. Search queries could also be useful for predicting sudden changes in consumer behaviour, which existing models find hard to anticipate.
