A great search engine understands a user’s intent and returns the most relevant results. Modern search engines, like Google, Bing, and DuckDuckGo, are really good at this. However, most search technologies fall short because they only look at the words in the query, not the intent behind it. They use “keyword search,” which has been around for more than two decades.
But keyword search has a major flaw: humans use different words to ask the same questions. Consider the following examples:
The users behind these searches have similar or the same intent, and a keyword-based approach will fail to provide the most relevant results for these searches that are looking for the same information.
We trained Google’s BERT to better understand what a customer is really looking for. This means that, unlike keyword-based systems, Answers knows that someone searching “send back shoes” is looking to kick off a return process, and that someone asking about a “dislocated shoulder” probably needs an orthopedist.
These vectors have 768 dimensions—which is impossible for us to wrap our heads around—but you can visualize the process in two dimensions here: