Microsoft is poised to add machine-reading results to Microsoft Search

Microsoft looks ready to commercialize more of its AI technology, this time in the form of new machine-reading comprehension capabilities built into Microsoft Search.

projectturing.jpg

Credit: ZDNet

For the past several years, Microsoft researchers have been focused on finding ways to make commercial use of machine-reading technology. It looks like some of that work is about to become commercialized in the form of bringing machine-reading comprehension into search results. Based on information in Microsoft's Ignite conference session list, Microsoft may be ready to show this off as soon as next week. 

Machine-reading comprehension involves the automatic understanding of text. It involves computer vision, natural-language understanding, and other technologies.

"In search applications, machine comprehension will give a precise answer rather than a URL that contains the answer somewhere within a lengthy web page. Moreover, machine comprehension models can understand specific knowledge embedded in articles that usually cover narrow and specific domains, where the search data that algorithms depend upon is sparse. In this session, see and learn about the latest innovation with natural language and machine reading comprehension in Microsoft Search," reads a blurb from one Ignite session slated for next week.

This technology is codenamed "Project Turing." (I found a mention of a Project Turing dating back to 2015, but it doesn't seem to be connected to Microsoft; it was a project of an Indian machine-learning startup called Snapshopr.) Microsoft's Project Turing involves an "AI-powered search for enterprise."

Search has been a big focus in the past few Ignite conferences. Last year at Ignite, Microsoft officials detailed how the company is unifying its search capability across Windows, its Edge browser and Bing. The unified search experience is called Microsoft Search

Microsoft officials have said previously that it's a combination of the Microsoft Graph, the company's centralized application programming interface, plus semantic knowledge from Bing, that powers the unified search experience. The Microsoft Graph is what contributes an understanding of users' work life, meaning the documents, the entities, the people they work with regularly and other everyday signals. Bing contributes an understanding of the world outside an organization, with acronym and entity extraction, machine reading comprehension and computer vision.  

Officials said the combination of these Microsoft Graph and Bing capabilities will allow Microsoft Search not just to answer simple queries, but more complex ones, as well, such as "Can I bring my partner and kids on a work trip?" by using machine reading comprehension coupled with an understanding of an organization's internal documents.  

In 2017 at the company's Faculty Research Summit, Microsoft officials showed off work it was doing in machine reading. Execs talked up efforts in creating a new neural network architecture in which Microsoft was investing that was called ReasoNet. 

That same year, Microsoft bought deep-learning startup Maluuba which also has focused heavily on machine reading. Maluuba pioneered ways to train machines to seek information and read and reason. The company also is finding ways to train machines to ask questions. Maluuba's machine-reading comprehension (MRC) system was able to ingest a 400-page auto manual and then answer user questions based on it in real-time.

Microsoft Research Montreal has been developing algorithms that can answer questions about new documents with a limited amount of training data. 

"Our primary goal is questioning-answering in the real world: we envision an experience where getting the answers you need to complex questions about your documents is simple, effective, and intuitive," according to the machine-reading comprehension page on Microsoft Research's site.

I asked Microsoft officials if they'd comment on what's happening with Project Turing and Microsoft Search, but didn't receive an answer.