Researchers and engineers in Microsoft Research (MSR) and Bing R&D have been working closely together and have made significant contributions to the science of search. These contributions often start as basic research ideas published as scientific articles in international conferences and journals. These articles are reviewed by search experts from industry and academic institutions around the world, and only those that significantly advance the state of the art are accepted and published. Many of the concepts in these articles from Microsoft become specific techniques for enhancing the Bing search engine positively impacting millions of users.
As an example, this week our researchers are participating in the IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR) 2012 (www.cvpr2012.org). CVPR is an international conference and an important event for the researcher and academic communities. Microsoft is presenting 41 scientific articles this week in CVPR that have already been reviewed by search experts covering a large number of topics such as ranking and image search.
In particular, the article Salient Object Detection for Searched Web Images via Global Saliency describes an innovative technique to detect the existence of objects for thumbnail images. Microsoft Research has described this article in some detail in this blog and the method has been shipped in the Bing image search engine. Other interesting CVPR articles from Microsoft that cover search topics in CVPR are Scalable k-NN graph construction for Visual Descriptors, Fast approximate k-means via cluster closures, and Image search results refinement via outlier detection using deep contexts.
There are multiple international conferences like CVPR every year that cover search and machine learning topics such as SIGIR, WSDM, WWW, and more. There are also multiple related scientific journals in the space such as Machine Learning, Journal of Machine Learning Research, IEEE Transactions on PAMI, to name a few. Often a new search technique is initially presented as an article at a conference and then it is published in more detail in a journal.
Microsoft is an active participant in the research community and publishes dozens of scientific papers every year that cover all aspects of search from information retrieval to machine learning, distributed systems, and human-computer interaction. Many of the best techniques in Bing don’t get disclosed publicly for competitive reasons but all the research that Microsoft publishes in the public forums above is related to our efforts to advance the state of the art in search science, which is important for the development of Bing. MSR and Bing researchers have long been leaders in publishing scientific articles and engaging with the research community in the search space.
As part of the Bing Search Quality Insights series, we will sporadically report on scientific events in industry and academia and will share links to some of the Microsoft search-related papers. Keep in mind that these articles are usually written by scientists and researchers and in general they are more complex and scientific compared with general tech blogs. I hope you enjoy these articles as much as we Bing engineers do. Happy reading!
– Harry Shum, PhD, Corporate Vice President, Bing R&D