How Microsoft's Bing-related research is funneling back into products
Microsoft officials have been attempting to make more explicit how Microsoft Research (MSR) projects end up benefiting the company on the commercial-product front. One way they've attempted to do this is by publishing lists of research technologies that end up "transferring" to product groups.
In looking at these lists recently, I was struck by how many MSR projects are influencing Bing.
Some of these, like the work by MSR to improve the accuracy of speech in Microsoft's cloud-based speech recognition systems, have shown up relatively quickly as new features in products like Windows Phone's Bing voice-search capability. And based on a new TechFest 2013 video clip uncovered by Microsoft watcher Stephen Chapman, it looks like the Windows Phone team is poised to adopt even more of the work from Microsoft researchers to improve the error rate and latency issues around Bing Mobile's voice search.
Bing Mobile voice search is just the tip of the transer iceberg. Look at the list of other Bing-related Microsoft Research projects that have moved to Microsoft's product groups in the past two years alone:
PingMesh: A full-mesh latency monitoring system that provides a key capability for deep understanding on end-to-end network latency characteristics crucial for Bing’s applications and services.
Dictionary: For those seeking to learn a foreign language, researchers created an approach to flash card–based learning in which news items are drilled into short-term memory for immediate prospective use.
Zozzle: A system that identifies malicious Web pages. Zozzle is designed to perform static analysis of JavaScript code on a given site and quickly determine whether the code is malicious and includes an exploit.
Engkoo: Microsoft Research Asia contributed a new feature for Engkoo to make it easier for ESL Chinese users to complete their English tasks in Bing.
PNav: An algorithm that predicts which Web search results a user will click for repeat queries.
Large-Scale Face Image Search: Incorporated into Bing multimedia search to index and show images from a search query.
ScopeStudio: Researchers collaborated on static code analysis that allows SCOPE developers to discover certain types of issues and defects earlier in the development cycle before actually submitting a job to a Cosmos cluster.
Bing Translator Widget: This feature evolved into a BingBar button for 40+ languages. The feature also lets people know when the BingBar can help with something being done online, like translating a word, sentence or article.
Bing Translation: Researchers teamed up with Windows Phone engineers to create the Translator app for Windows Phone that uses the same translation technology powering the Bing translator to help people understand languages by translating text, audio, and video inputs
Bing Dictionary: In collaboration with both Bing and Office teams, researchers developed the Bing English Assistance feature -- an English monolingual dictionary -- that is available for English language versions of Office 2013.
TV (Leibniz): Researchers leveraged machine learning algorithms to create the ability to sync technology, product, and service for matching movies and TV shows/episodes. Leibniz was first deployed in Bing as the basis for Entity Actions, making Bing a task engine, starting with automatic tasks such as renting and streaming movies and TV shows.
All this is just another reminder to those who think Microsoft might be gearing up to sell off Bing that this is pretty unlikely. Search seems to see it as an area in which the company is going to continue to invest.
And while we're talking Bing, the Bing team added more features to its Snapshot "Answers at a Glance" feature this week. Microsoft originally rolled out Snapshot for Bing last June.