The Magic That Allows Google To Understand Our Questions
March 30, 2021Many of us take Google’s Search engine for granted. Despite it being around for more than 2 decades, many of us do not appreciate the technology behind this amazing functionality. Search is the backbone of our internet search queries. It helps us find what we are looking for whether its a specific cat video or an explanation on nuclear fission.
Google’s clever algorithms constantly trawl the internet, searching, indexing and archiving important data to make it easier for us to find what we are looking for. This technology plays such a vital role in our lives that the term “Google It” is globally accepted as a synonymous way of referring to the act of searching for something online.
But, for Search to accurately find what we are looking for, it has to first understand what we’ve typed into its search bar. That in itself is a major challenge, according to Google. We basically make one of two types of errors when incorrectly searching for something. The first is typing our best estimate of what we want and the other is misspelling a word.
Though we might not consider it a big deal, it is actually a major problem. Despite how powerful Google Search has become, it still doesn’t have the ability to read minds. And as such, when we incorrectly type a search query or misspell a word, the algorithm has to come up with a best estimate as to what it thinks that we want.
For example, if you are searching for the “best hotels on the east coat“, Google has to understand the context of the sentence, figure out the mistake, then present you with an accurate search result. The same is true of misspelled words like “outube” or “tourube” when we want YouTube. Its algorithm tries to predict what we want by considering the proximity of the keys pressed to what we intended to press.
This is all possible because of Google’s deep neural net. It’s a new spelling algorithm that was introduced last year. It allows google to run a model with more than 680 million parameters in under two milliseconds. That’s why Search is quick to respond to what we’ve typed. The same is true of its natural language understanding models.
This incredibly clever backend system is why Google’s Search functions are extremely quick at responding fairly accurately to our search queries. But if it isn’t quite sure of what we want, it will display its best estimate by stating “did you mean……” But it will still give you the option of searching for what you originally typed.
As such, the next time you look something up on Google, take a few moments to admire the magic that is taking place behind the scenes that enable us to find what we want. It’s astonishing as to how much AI has to do, to understand that we wanted to look for YouTube instead of gibberish like “uobyoy”.