In line with its CEO Satya Nadella’s affirmation to democratize artificial intelligence (AI) Microsoft today released its Space Partition Tree and Graph (SPTAG) algorithm that drives Bing search as a MIT licenced open code on GitHub.
Announcing this on its official blog, Microsoft said the algorithm allowed users to take advantage of the intelligence from deep learning models to search through billions of pieces of information, called vectors, in milliseconds. That, in turn, meant they could deliver more relevant results to users ever so quickly. Effectively, it is called vector search.
How does all this matter to the average Web surfer?
As we are all aware, Online search has evolved rapidly over the last few years. It was a simple process in the early days. Key in a few words & go through thousands of pages of search results. No longer so.
Today users have options to search – they click a picture on their smartphone & drop it into a search box, wanting more details? Or they may simply ask an intelligent assistant to find something. These tasks pose a challenge to traditional search engines, which, said Microsoft, “are based around an inverted index system that relies on keyword matches to produce results.”
“Keyword search algorithms just fail when people ask a question or take a picture and ask the search engine, ‘What is this?’” said Rangan Majumder, group program manager on Microsoft’s Bing search and AI team, in the blog post.
So, as an example, if you were to ask how tall is the tower in Paris, without even mentioning the word, ” Effel”, Bing, using vector search, can now give you the exact result in natural language.
With Bing search, said Microsoft, the “vectorizing effort” has extended to over 150 billion pieces of data indexed by the search engine to bring improvement over traditional keyword matching. These include single words, characters, Web page snippets, full queries & other media. Once a user searches, Bing can scan the indexed vectors & deliver the best match.
Vector assignment is also trained using deep learning technology for ongoing improvement. The models consider inputs like end-user clicks after a search to get better at understanding the meaning of that search.
For the technically inclined, here’s the post.
Image Credit: Microsoft
