Google does a lot of things. It seems it to grow a new hand to stick in yet another pot every six months or so. However, the company's first product (and arguably its most important one), goes by the name of Google Search. According to Google, more than 1000 years of labor have gone into building the Google Search algorithm. The company is constantly tweaking this but we don't really get a lot of insight into that process.
This week, Google is offering us a glimpse of how search works. The company has, for the first time, posted an uncut video documenting one of its "Quality Launch Review" meetings. Google holds the meeting on Thursdays to discuss possible algorithmic improvements and make decisions about what to launch. This meeting in particular covered improving spelling suggestions for searches with more than 10 words. It impacts just 0.1 percent of Google's traffic but even for small details like this one, aspects such as relevance, spam, latency, cost, and language impact are taken into account.
Check out the video below for a taste of what it's like to be a part of Google's Search team:
Jane McEntegart is a writer, editor, and marketing communications professional with 17 years of experience in the technology industry. She has written about a wide range of technology topics, including smartphones, tablets, and game consoles. Her articles have been published in Tom's Guide, Tom's Hardware, MobileSyrup, and Edge Up.
I think it's more that they've cached everything you've ever done and then some since their beginning, across all services, and then cross referenced all that information creating a unique yet robust algorithm that is specifically targeted based off the general category you fall under as a Netizen.
robochumpCoulda swore the secret was a Foxconn warehouse full of pre-teen Chinese kids slaving away to answer Google search requests. Disappointed....your thinking of isearch
It's .1% of queries and they're wasting the time of dozens of people trying to fix it? I'd have shelved it off the bat with a memo saying "not statistically significant enough to warrant resource allocation towards solution".
