Wednesday, February 24, 2010

Google and it's Approach to Search

Wednesday, February 24, 2010
I'm an avid reader of wired.com; recently, they posted an article on google and went into a never-before reported depth on the inner workings of the algorithm and how Google actually interprets your query and comes up with the most relevant sources. The full article can be found here: http://www.wired.com/magazine/2010/02/ff_google_algorithm.

What I found interesting was how intelligent the algorithm actually was. Google started off with the first breakthrough with having the search engine look at the popularity of sites by looking at how many links pointed to them. Later it moved on to do such things as even learn context and underlying meanings of words. For example it has been built to learn human-made synonyms. For instance, they talk in Wired about the meaning of dog/puppies. A puppy and a dog are two distinct words, but google uses the fact that when users search up a term such as pictures of dogs, they go back, delete the term dog and replace it with pictures of puppies. Google uses this to learn that to human beings dog and puppy are in some ways similar. This doesn't seem as difficult but Wired goes on to talk about the association between boiling water and hot. Google has learned enough that it can also seperate a hot dog from a boiling dog. In addition, it understands that a hot dog isn't a dog that is very hot, but rather a type of food. Similarly when you search up "strcraft", google has learned that this is a common misspelling for starcraft and tells you to change your search term. Google has been built in such a way that everytime a user enters a search term, it's learning from the user associations between words and the meanings of words. The last example I'll use shows it's learning of word associations. For example, the search engine understand the word horse to be an animal and white to be a color. However, when you pair up the two White Horse (even if you leave a space between them), the search engine recognizes that it no longer has it's old two meanings but is instead the name of a city (and incidentally a song by Taylor Swift lol). Additionally it's also realized the difference between the order of the words. Typing in horse white will give you similar results to before, but one of the top choices now talks about horses whose color is white.

What is interesting is the evolution of computers. The first computers were really basic calculators. Anything even slightly creative or different than what they have seen before would bring up errors. However, in the past few decades, algorithms have gone on to achieve much more complexity, I would daresay, some extent of intelligence. Google essentially has developed a capacity to think and even to LEARN! It seems that movies like Terminator and AI aren't so far-fetched after all...

0 comments:

Post a Comment