Google Releases Tool to Show Off Natural Language Search Capabilities by @MattGSouthern
Google’s research division has released a new tech demo of sorts, which shows off advancements the company has made toward understanding natural language.
The new search experience, called “Talk to Books,” invites users to ask questions which Google will respond to with passages from books. Google claims the tool does not rely at all on keyword matching.
I tested it out using the number one question every SEO wants an answer to: “How do I rank first in Google?”
In this example Google returned a variety of results from marketing textbooks and, as advertised, the results appear to be surfaced based on context — not keyword matching.
Talk to Books works by searching through all the sentences in over 100,000 books to find passages most relevant to the query based on semantic meaning.
”Natural language understanding has evolved substantially in the past few years, in part due to the development of word vectors that enable algorithms to learn about the relationships between words, based on examples of actual language usage. These vector models map semantically similar phrases to nearby points based on equivalence, similarity or relatedness of ideas and language.”
Google says the models driving this experience were trained on a billion conversation-like pairs of sentences. There are no predefined rules bounding the relationship between queries and results.
For the time being, this search technology is only being used to find information in books. Although one can’t help but wonder what the SEO landscape might look like if this tech was applied to regular search results.